- Email: [email protected]
Characterization of orbital masses by multiparametric MRI
Accepted Manuscript Title: Characterization of orbital masses by multiparametric MRI Author: Sa-Ra Ro Patrick Asbach Eberhard Siebert Eckart Bertelmann Bernd Hamm Katharina Erb-Eigner PII: DOI: Reference:
S0720-048X(15)30184-4 http://dx.doi.org/doi:10.1016/j.ejrad.2015.11.041 EURR 7331
To appear in:
European Journal of Radiology
Received date: Revised date: Accepted date:
4-8-2015 25-11-2015 30-11-2015
Please cite this article as: Ro Sa-Ra, Asbach Patrick, Siebert Eberhard, Bertelmann Eckart, Hamm Bernd, Erb-Eigner Katharina.Characterization of orbital masses by multiparametric MRI.European Journal of Radiology http://dx.doi.org/10.1016/j.ejrad.2015.11.041 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
Characterization of orbital masses by multiparametric MRI
Sa-Ra Roa, Patrick Asbach, MDa, Eberhard Siebert, MDb,
Eckart Bertelmann, MDc,
Bernd Hamm, MDa, Katharina Erb-Eigner, MDa
a
Department of Radiology, Charité - Universitätsmedizin Berlin, Campus Benjamin Franklin,
Hindenburgdamm 30, 12203 Berlin, Germany b
Department of Neuroradiology, Charité - Universitätsmedizin Berlin, Campus Charité Mitte,
Charitéplatz 1, 10117 Berlin, Germany c
Department of Ophthalmology, Charité - Universitätsmedizin Berlin, Campus Virchow-
Klinikum, Augustenburger Platz 1, 13353 Berlin, Germany
Corresponding author: Patrick Asbach, MD Department of Radiology Charité - Universitätsmedizin Berlin, Campus Benjamin Franklin Hindenburgdamm 30, 12203 Berlin, Germany Phone: +49 30 8445 3041 Fax: +49 30 450 7 527 953 E-mail address: [email protected]
2 Abstract Objectives DWI and dynamic contrast enhanced (DCE) MR imaging are techniques that allow insight to tumor vascularity and cellularity. We evaluated the diagnostic performance of multiparametric MRI (mp-MRI) in distinguishing benign from malignant orbital masses using standard anatomic imaging (sAI), DWI and DCE. Materials and Methods This prospective IRB approved study with written informed consent included 65 patients. mpMRI at 3 Tesla including DWI and DCE was performed in all patients. Parametric maps were generated for obtaining the perfusion parameters including Ktrans, kep, ve and iAUC and timesignal intensity curves were recorded to determine the curve pattern. Two radiologists rated the likelihood of malignancy on a five-point scale in three separate, randomized reading sessions (initially only sAI, afterwards sAI + either DWI or DCE and finally sAI + DWI + DCE). Data was statistically analyzed. Results 33 patients had malignant orbital masses and 32 patients had benign orbital masses (reference standard histopathology in 35 cases and clinical follow-up in 30 patients). The mean ADC of malignant masses differed significantly from the mean (SD) ADC of benign masses (0.825 [0.437] x 10-3mm²/s and 1.257 [0.576] × 10-3mm²/s, respectively) (P=0.001). Ktrans, kep and iAUC were significantly higher in malignant masses (P<0.01). The reading of sAI only resulted in a moderate specificity but poor sensitivity in differentiating benign from malignant lesions. Adding DWI and DCE images improved specificity and sensitivity considerably, being the highest for the combined reading of all sequences. Conclusion mp-MRI is a helpful tool in differentiating malignant orbital lesions from benign masses and should therefore be included in the routine diagnostic protocol for orbital imaging.
Keywords Orbital masses; Orbital MRI; Multiparametric MRI; Diffusion weighted imaging; Dynamic contrast enhanced imaging
3 Introduction
Orbital and intraocular masses are relatively uncommon compared to other mass lesions of the body. It has been described that two-thirds of orbital tumors are benign and one third is malignant (1). Characterization of orbital masses is crucial in the therapeutic strategy planning (2) as the patient management differs greatly depending on the dignity of the orbital mass (3). However, it is often difficult to differentiate malignant orbital masses from benign ones due to their similar clinical presentation with proptosis as the most common symptom (4). Magnetic resonance imaging (MRI) may help finding the diagnosis as there are some pathognomonic features for particular masses as for instance the cavernous hemangioma (5). But for cases without these pathognomonic characteristics it remains difficult to deliver a diagnosis based on MR imaging features because there are often unspecific and overlapping imaging findings (5). Aggravating this situation, rare tumor entities are unexpected and therefore may be misdiagnosed (5). It has even been proposed that none of the orbital imaging features including CT and MR imaging features had sufficient sensitivity to distinguish between malignant and benign orbital masses (6). Published data indicate that the use of advanced MRI sequence techniques like diffusion weighted imaging (DWI) with quantitative apparent diffusion coefficient (ADC) mapping and dynamic contrast-enhanced (DCE) MR imaging may provide additional information about the dignity and entity of orbital masses (5, 7-14). Based on the diffusivity of water in tissue DWI exploits the fact that the movement of water is normally restricted in malignant tumors, whereas DCE offers information on the rate of uptake and clearance of contrast media and can be analyzed to derive information on tissue and tumor vascularity. Tofts et al. defines a compartment model to describe abnormal capillary leakage and uses the following parameters: Ktrans = volume transfer constant between blood plasma and extravascular extracellular space (EES) in abnormal tissue, kep = efflux rate constant from EES to blood
4 plasma, ve = EES volume per unit volume of tissue and iAUC = initial area under the gadolinium concentration curve during the first 60 seconds (15).. Taking advantage of either of these, DWI and DCE, there are several research groups who have investigated multiparametric magnetic resonance imaging (mp-MRI) of the orbital cavity by adding one functional parameter such as DWI or DCE to the standard morphological sequences (4, 5, 7, 17, 18). There were suggestions that DCE or DWI may produce complementary information (19). However, to our knowledge, the diagnostic performance of combining conventional magnetic resonance imaging with DWI and DCE has not been investigated. The purpose of our study was to prospectively evaluate the diagnostic performance of mpMRI in distinguishing benign from malignant orbital masses using DWI and DCE MR imaging.
5 Material and Methods
Patients The institutional review board approved this prospective single-institution clinical study. Written informed consent was obtained from all patients. Between May 2013 and July 2014 a total of 91 patients with known or suspected orbital space-occupying lesion were recruited in the outpatient ophthalmology clinic. This study was registered at clinicaltrials.gov under ClinicalTrials.gov-ID (blinded for review). A senior ophthalmologist who is specialized on orbital tumors examined all patients before the MRI examination. Patients were excluded if they were not at least 18 years old (n=1) or if they had undergone therapy between histopathology and MRI examination (n=3). Other exclusion criteria were contraindications to MRI or contrast agent applications (GFR<30 ml/min; allergies), pregnancy and claustrophobia. Ultimately 87 patients were enrolled in this study and underwent MR imaging with standard anatomical imaging (sAI) including T1-weighted, T2-weighted and post-contrast T1-weighted sequences and multiparametric MR imaging (mpMRI) including diffusion weighted imaging (DWI) with ADC mapping and dynamic contrast enhanced (DCE) imaging. In addition to the previously mentioned criteria, nine patients were excluded because there was no mass lesion seen on MRI, five were excluded because the lesion was too small (≤ 4 mm) and partial volume averaging effects hinder accurate analysis on mp-MRI, another four patients had an incomplete index test and four patients had no reference standard, resulting in a final cohort with 65 patients. Final diagnosis was established through histopathological results from surgery (n=2) or biopsy (n=33), pathognomonic imaging findings (n=7) or clinical follow up over a period of a minimum of 6 months (benign lesions with no change, n=4), ophthalmoscopy (uveal
6 melanoma, n=17) or unequivocal clinical diagnosis (multi-organ metastatic breast cancer, n=1; pseudotumor orbitae, n=1).
MR imaging technique Imaging was performed on a 3 Tesla scanner (MAGNETOM Skyra; Siemens Healthcare, Erlangen, Germany) with either a single-channel 7-cm loop surface coil (for sAI) or a 20channel head coil (for DWI and DCE). The respective sequence parameters of the MR protocol are summarized in Table 1. Contrast medium was administered intravenously body weight adapted (0.1 mmol gadoteric acid/kg body weight) after the 3rd DCE measurement (approximately 13 seconds) as bolus injection in an antecubital vein with an injection speed of 2ml/s using a 0.60x25mm tube.
Image analysis Quantitative Analysis The sAI were used to measure the diameter of the lesion in three dimensions. The mean diameter was calculated. In the DW images a freehand drawn ROI was placed in the lesion at b = 0 and 1000 sec/mm², while the edges of each lesion were avoided to minimize the effect of partial volume averaging. Similarly, an identical ROI was placed within the vitreous humor. The number of pixels, the size of the ROI and the SI at b = 0 and 1000 sec/mm² were recorded for each patient. The ADC values of the respective ROIs were calculated according to the formula
ADC=
.
A 2-ADC threshold model was used with 0.8 x 10-3mm²/s and 1.2 x 10-3mm²/s as thresholds to categorize lesions in three groups.
7 For the image analysis of the DCE MR images we used a dedicated postprocessing software program (Tissue 4D; Siemens Healthcare) which supplies pharmacokinetic calculation on a pixel-by-pixel basis using a 2-compartment model based on the Tofts model. Concerning the arterial input function it was decided to choose the intermediate sampling method. A freehand ROI was drawn over the area of maximum DCE MR imaging signal within each lesion and voxelwise perfusion maps containing the following parameters were generated: mean Ktrans, kep, ve and iAUC. Using the same postprocessing software program we recorded time-signal intensity curves for each lesion in addition to the arrival time, the time of the peak and the last time point of the contrast agent and the respective signal intensities. There were three different types of curve pattern. Type 1 curves showed a persistent pattern (i.e., SIlast time = SIpeak time or SIlast >= SIpeak), type 2 curves showed a plateau pattern (i.e., SIpeak time-SIlast time ≤10% of SIpeak time) and type 3 curves showed a washout pattern (i.e., SIpeak time-SIlast time>10% of SIpeak time). T1-mapping was included in the protocol but we refrained from ROI analysis since no standard of reference parameter have been established yet.
Qualitative Analysis Two radiologists with 11 and 8 years of experience in MR imaging who were blinded to the diagnosis, clinical history and histopathologic findings independently reviewed the MR images in randomized order in three reading sessions, with a three week interval between the first two sessions and a six months interval between the second and third reading session. Prior to the first reading the readers underwent a short training session that displayed five benign and five malignant orbital lesions in the respective sequences. In the first session each reader reviewed the sAI sequences (T1-weighted, T2-weighted and T1 post-contrast) only, followed by the addition of DW images and the corresponding ADC maps. For the 2nd session the readers reviewed initially the sAI sequences again followed by the DCE MR
8 images. For the 3rd session the readers reviewed the sAI sequences together with the DWI and DCE MR images. At each reading the readers were instructed to score the images for malignancy on a 1 to 5 Likert scale to convert subjective assessments into quantifiable parameters (1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant). They also independently graded the sequence quality of the standard anatomical images and the mp-MR images (DCE and DWI) with regard to artifacts on a scale of 1 to 5 (‘The study shows a perfect representation of the orbit without artifacts’, 1 = strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree and 5 = strongly agree).
Statistical analysis Descriptive statistics for the outcome of the qualitative analysis compared to the reference standard were performed for each reader and each sequence and the corresponding sensitivity and specificity were determined. Differences between the sensitivities and specificities of each viewing setting were tested for significance based on the binomial test. The nonparametric Mann-Whitney-U test for unpaired variables was applied to determine whether the diameter, ve, ADC, Ktrans, kep and iAUC of benign lesions differed significantly from those of malignant lesions using the Statistical Package for the Social Sciences software (SPSS version 11.5). Using the Pearson Chi² Test the curve pattern was tested for significance. The software 'R' (R version 3.1.2) was used for comparison of ROC curves and to test inter-reader and intra-reader variability (Cohen’s kappa test). Since there were multiple comparisons Bonferroni adjustment of P values was used. differences.
P < 0.01 indicated statistically significant
9 Results
Patients Out of 91 eligible patients 87 patients underwent MR imaging, and 22 were excluded (Fig 1). The final cohort included 65 patients. In 32 cases (32/65, 49%) malignancy was confirmed, of whom there were 17 men (age range, 26-80 years; mean age, 60 years) and 15 women (age range, 38-68 years; mean age, 55 years). The other 33 patients (33/65, 51%) had benign lesions, of whom there were 20 men (age range, 20-75 years; mean age, 53 years) and 13 women (age range, 29-88 years; mean age, 61 years). The 33 malignant lesions included the following diagnoses: conjunctival carcinoma (n=1), squamous epithelial carcinoma (n=1), adenoid cystic carcinoma (n=1), not otherwise specified carcinoma but histopathology states it as malignant (n=1), follicular lymphoma (n=2), marginal zone lymphoma (n=4), mantle cell lymphoma (n=2), esthesioneuroblastoma (n=1), uveal melanoma (n=17), melanoma (n=1), metastasis of breast cancer (n=2) of whom one was proven by histopathology while the other one was confirmed by follow up (within three month since the last MR examination a newly developed breast cancer was found and the patient developed multi organ metastatic disease).
Except for uveal melanoma where
fundoscopy is the gold standard for diagnosis (20) and the above-mentioned patient who developed metastasis from breast cancer all other malignant lesions were proven by histopathology. The diagnoses of the 32 benign lesions were as follows: hemangioma (n=5), meningioma (n=3), angiomatosis (n=1), neurofibroma (n=1), lipoma (n=2), lymphatic tissue (n=2), lymphangioma (n=1), sarcoidosis (n=1), endocrine orbitopathy (n=1), inflammation of lacrimal gland (n=3), amyloidosis (n=1), hematoma (n=1), pseudotumor (n=2), mucocele (n=1), cyst (n=1), schwannoma (n=1), orbital vein thrombosis (n=1), granulation tissue (n=1), myxoma (n=1), pyocele (n=1), not otherwise specified lesion but histopathology claims it as
10 benign (n=1). There were pathognomonic imaging findings for four hemangiomas, two optic nerve sheath meningiomas and the orbital vein thrombosis. The pyocele and one pseudotumor were proven by surgery. The other pseudotumor was determined as definite clinical diagnosis. The angiomatosis, lymphangioma, endocrine orbitopathy and hematoma were confirmed by follow-up. The remaining 18 cases were proven by histopathology.
Quantitative analysis The mean diameter (SD) of benign lesions was 1.74 (0.86) cm (range 0.47 – 5.19 cm) and was 1.64 (0.91) cm (range 0.72 – 4.59 cm) of malignant lesions. The difference was not statistically significant (p=0.259). The mean (SD) number of pixel per ROI in the DW images was 29 (17). The mean (SD) size of a ROI was 29 (20) mm². The mean (SD) ADC of the vitreous humor was 2.686 (0.257) x 10-3mm²/s. There were significant differences regarding the ADC between malignant and benign lesions (p=0.001). In the malignant lesions the mean (SD) ADC was 0.825 (0.437) x 10-3mm²/s, while the mean (SD) ADC of benign lesions was 1.258 (0.577) x 10-3mm²/s. Applying an ADC value of 1.0 x 10-3mm²/s as a threshold for differentiation between malignant and benign lesions resulted in 75% of the malignant studies to be found below the threshold and 25% above the threshold. The mean (SD) ADC of the ocular melanoma subgroup (n=17 of 33 malignant lesions) was 0.771 (0.016) x 10-3mm²/s. ROC analysis revealed that a ADC of less than 8.141 x 10-3mm²/s was more than 88% specific for malignancy, and a ADC of less than 1.151 x 10-3mm²/s was more than 88% sensitive for malignancy (AUC=0.767). For a more accurate approach a 2-ADC threshold model was used with 0.8 x 10-3mm²/s and 1.2 x 10-3mm²/s as thresholds. As a result only 12% of the malignant lesions had an ADC above 1.2 x 10-3mm²/s (Fig 1). For the benign lesions the same thresholds
11 were used as for the malignant lesions. Again the 2-ADC threshold model reduced the number of benign studies, which wrongly indicated malignancy, from 34% to 12.5%. The exact numbers for the DCE perfusion parameters are given in Figure 1. In 3 studies the patients underwent 2D Perfusion imaging and were therefore excluded from the statistical evaluation of the perfusion parameters. In the remaining 62 studies the perfusion parameters Ktrans, kep and iAUC were significantly higher in malignant lesions (p<0.01). ROC analysis suggested that a Ktrans of more than 0.257 was more than 90% specific for malignancy, and a Ktrans of more than 0.077 was more than 90% sensitive for malignancy (AUC=0.781). A kep of more than 1.206 was more than 90% specific for malignancy, and a kep of more than 0.617 was more than 90% sensitive for malignancy (AUC=0.846). Also, an iAUC of more than 14.468 was more than 90% specific for malignancy, and an iAUC of more than 4.635 was more than 90% sensitive for malignancy (AUC=0.735). On the contrary the mean ve in malignant lesions was lower than in benign lesions and there were no significant differences for ve (p=0.108). The curve pattern could be obtained for all of the 65 patients. The curve pattern for malignant lesions differed significantly from benign lesions (p<0.01). Malignant lesions mainly (25/33) showed a washout pattern and benign lesions mostly (21/32) a persistent pattern.
Qualitative analysis Since the sAI images were scored twice in two independent sessions there were a total of 130 sAI readings. While for the combination of the standard anatomic images with DW images and with DCE MR images there were 65 studies each. As shown in Figure 2 both readers rated the images mostly as indeterminate or as probably malignant if they had only the standard anatomic images at their disposal. The combination of the standard anatomic images and DW images proved to be most conclusive to decide whether malignancy was detected on the images. While the addition of DCE MR images to the standard anatomic images was
12 inferior compared to DWI for determining malignancy, the combination of both, adding DW and DCE images to sAI produced the best results to differentiate between malignant and benign lesions. The precise details of the ratings and the respective sensitivity and specificity for each reader and each image-viewing-setting are given in table 2. Figure 3-7 gives examples of the sAI images, the DWI, the ADC-map, the post-contrast T1-weighted images and the time-intensity curve of benign and malignant lesions. For reader 1 the standard anatomic images in combination with the DW images yielded the highest sensitivity (71.87%) and the highest specificity (96.96%). For reader 2 the sensitivity was highest with the combination of standard anatomic images and DCE images (71.87%), while the highest specificity was found with the standard anatomic images in addition to the DW images (100%). All in all both the addition of DW MR images for diagnostic purposes and the addition of both, the DW and DCE MR images produced statistically significant improvement regarding sensitivity and specificity for both readers (P<0.01). ROC analyses of the readings of the two readers are displayed in Figure 9. Comparing ROC curves the addition of DW MR images to sAI images improved the AUC significantly (Reader 1: p=0.015; Reader2: p=0.018). Also, the AUC results of the reading using sAI, DW and DCE MR images simultaneously differed significantly from the AUC results using sAI images only (Reader 1: p=0.003; Reader 2: p=0.006). The differences between the first and second sAI readings of the first reader, as well as the second reader indicated a different level of confidence (intra-observer variability Reader 1: κ=0.227 ‘fair agreement’; Reader 2: κ=0.582 ‘moderate agreement’). The inter-observer variability between Reader 1 and Reader 2 decreased by the addition of DCE and DWI images (first sAI reading: κ=0.292 ‚fair agreement‘; second sAI reading: κ=0.494‚ moderate agreement‘; sAI+DWI reading: κ=0.534 ‚moderate agreement‘; sAI+DCE reading: κ=0.627 ‚substantial agreement‘; sAI+DWI+DCE reading: κ=0.644 ‚substantial agreement‘).
13 Both readers rated that the standard anatomic images and the DCE MR images showed a perfect delineation of the structures of the orbital cavity in 64 of the 65 studies (score of 4 or 5). In one case they rated the sequence quality with regard to artifacts as 3. Reasonable diagnostic image quality with DW imaging was obtained in 58 of the 65 studies for reader 1. In the remaining studies reader 1 rated 3 in 5 studies and 2 in 2 studies. Reader 2 rated in 51 studies 4 or 5, in 9 studies 3, in 4 studies 2 and in one case 1.
14 Discussion
Aim of the study was to demonstrate significant quantifiable differences between malignant and benign orbital masses using mp-MRI: The lower the ADC value of an orbital mass, the more likely the mass was malignant, while benign orbital masses showed significantly higher ADC values. Supplementary, high values of Ktrans, kep and iAUC revealed a high probability for malignancy. Investigating the ve values among the orbital masses a tendency was observed: Benign masses demonstrated higher values while malignant lesions more often presented lower values. Furthermore our results propose that DWI and DCE MR imaging are complimentary: For instance one patient with sarcoid exhibited an ADC value that could be interpreted as either benign or malignant which may be due to the fact that the dominant reaction is a dense fibrous tissue with restricted diffusion. However when investigated with DCE MR imaging the lesion had a persistent curve pattern and low values for Ktrans, kep and iAUC. This way the lesion could be rightly categorized as benign. Similar to Yuan et al. most of the malignant lesions showed a washout pattern (curve type III; 25 of 33) and most of the benign lesions showed a persistent pattern (curve type I; 21 of 32) (17). DWI visualizes differences in tissue cellularity and has been used throughout the body to categorize lesions in 'benign' or 'malignant' (21). According to their degree of cellularity masses can be distinguished in malignant and benign. High cellularity of a tissue causes high restriction of water molecules in this tissue, which in turn is associated with lower ADC values. Sepahdari et al. described that with the addition of DWI, improved accuracy of MR imaging in the characterization of indeterminate orbital lesions may help triage patients to either early intervention or initial conservative treatments (9). In our study using a 2-ADCtreshold model had better results than using only one threshold for distinguishing between malignant and benign masses (for example: malignant masses which simulate benign masses:
15 1 ADC- Threshold: 24.24%, 2-ADC-Thresholds: 12.12%). This observation corresponds to other findings of Sepahdari et al. who described a 2-ADC threshold model to categorize orbital masses with high confidence (8). Fatima et al. reported a mean ADC value of 0.77 x 10-3mm²/s for malignant lesions and a mean ADC for benign lesions 1.23 x 10-3mm²/s (10). Sepahdari et al. reported a mean ADC value of 1.43 ± 0.41 x 10-3mm²/s for malignant lesions and a mean ADC for benign lesions of 0.90 ± 0.36 x 10-3mm²/s (8). In our study, the mean (SD) ADC was 0.825 (0.437) x 10-3mm²/s for malignant lesions, while the mean (SD) ADC of benign lesions was 1.258 (0.577) x 10-3mm²/s. ROC analysis revealed that an ADC of less than 8.141 x 10-3mm²/s was more than 88% specific (and less than 6.456 x 10-3mm²/s was more than 90% specific) for malignancy, and a ADC of less than 1.151 x 10-3mm²/s was more than 88% sensitive (and less than 1.462 x 10-3mm²/s was more than 90% sensitive) for malignancy. It was noticeable that with only one cut- off value, the rate of clearly erroneous categorized lesions was significantly higher than with 2 cut-off values. For clinical practice, this means, that 8 malignant lesions would have been categorized as benign with only one cutoff value and 4 malignant lesions with 2 cut-off values, which would have a more hesitant therapy approach as a result. The introduction of a 2-ADC threshold model would mean that in the 6 borderline cases further diagnostic work-up would be used to determine the dignity of the lesion and the probability is increased to inflict upon the correct treatment of each lesion. Another noninvasive advanced MRI sequence is the DCE MR imaging. DCE MRI visualizes tumor vascularity pattern and has been used to characterize various tumors throughout the body (22). Our study showed that analysis of the curve type (persistent, plateau and washout pattern) allowed a confident characterization of the respective lesion into 'benign' or 'malignant'. There were only 2 malignant lesions that were classified as benign, whilst with DWI 8 (ADC>1.0 x 10-3mm²/s) or 4 malignant lesions (ADC>1.2 x 10-3mm²/s) were classified wrongly as benign. The analysis of objective DCE parameters (Ktrans, kep and iAUC)
16 was significantly higher in malignant lesions than in benign ones. Higher vascular permeability and higher perfusion is typically observed in malignant lesions (23). The results of the qualitative analysis suggest that the visual impression of DWI and DCE images increases the confidence of the reader. The reading of native images only resulted in a moderate specificity and an unsatisfying sensitivity. Adding both, DWI and DCE images improved specificity and sensitivity considerably, where the DWI showed better results increasing confidence of both readers. There are already studies that have dealt with either DWI or DCE, but to our knowledge there is no study that has tested both imaging techniques on one group of patients. Therefore, selection bias that may have been occurred in studies that dealt with DWI or DCE only could be avoided in our study. Furthermore, our study considered qualitative image analysis as well, whilst many studies are restricting on quantitative analysis only. Our study has several limitations. A patient population bias cannot be excluded which is reflected in the distribution of the orbital masses regarding their dignity. The tumors included reflect the referral pattern to our orbital clinic, which included equally benign and malignant lesions, and not 2 third to 1 third. Another limitation may be the manual delineation of lesions and calculation of DCE parameters over the entire tumor. The SI of the ROI is averaged and peak enhancement value is influenced by the complexity and heterogeneity (e.g. necrosis) of the mass. Furthermore, no multi-channel small loop surface coil was available, although the DWI and DCE sequences were acquired with the multi-channel head coil covering both orbits. In conclusion, this prospective study confirmed that mp-MRI through adding DWI or both DWI and DCE MR imaging improved not only readers’ confidence in categorizing orbital mass lesions (qualitative analysis) but also showed a good diagnostic performance in terms of quantifiable results (quantitative analysis). DWI and DCE MR imaging are helpful tools in differentiating malignant orbital lesions from benign masses. We therefore recommend including DWI and DCE MR imaging in the routine diagnostic approach of orbital masses.
17 Conflict of interest The authors have no conflict of interest to disclose.
AUTHOR DECLARATION We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing we confirm that we have followed the regulations of our institutions concerning intellectual property. We further confirm that any aspect of the work covered in this manuscript that has involved human patients has been conducted with the ethical approval of all relevant bodies and that such approvals are acknowledged within the manuscript. We understand that the Corresponding Author is the sole contact for the Editorial process (including Editorial Manager and direct communications with the office). He is responsible for communicating with the other authors about progress, submissions of revisions and final approval of proofs. Patrick Asbach, MD (Corresponding Author) [email protected] Signed by all authors as follows:
Sa-Ra Ro Patrick Asbach Eberhard Siebert Eckart Bertelmann Bernd Hamm Katharina Erb-Eigner 31/07/2015
Acknowledgments K. Erb-Eigner is supported by the ‘Rahel Hirsch Program’ funded by the Charité – Universitätsmedizin Berlin.
18 References
1. Shields JA, Shields CL, Scartozzi R. Survey of 1264 patients with orbital tumors and simulating lesions: The 2002 Montgomery Lecture, part 1. Ophthalmology. 2004;111(5):9971008. 2. Lemke AJ, Kazi I, Felix R. Magnetic resonance imaging of orbital tumors. European radiology. 2006;16(10):2207-19. 3. Valvassori GE, Sabnis SS, Mafee RF, Brown MS, Putterman A. Imaging of orbital lymphoproliferative disorders. Radiologic clinics of North America. 1999;37(1):135-50, x-xi. 4. Razek AA, Elkhamary S, Mousa A. Differentiation between benign and malignant orbital tumors at 3-T diffusion MR-imaging. Neuroradiology. 2011;53(7):517-22. 5. Xian J, Zhang Z, Wang Z, Li J, Yang B, Chen Q, et al. Evaluation of MR imaging findings differentiating cavernous haemangiomas from schwannomas in the orbit. European radiology. 2010;20(9):2221-8. 6. Ben Simon GJ, Annunziata CC, Fink J, Villablanca P, McCann JD, Goldberg RA. Rethinking orbital imaging establishing guidelines for interpreting orbital imaging studies and evaluating their predictive value in patients with orbital tumors. Ophthalmology. 2005;112(12):2196-207. 7. Roshdy N, Shahin M, Kishk H, Ghanem AA, El-Khouly S, Mousa A, et al. MRI in diagnosis of orbital masses. Current eye research. 2010;35(11):986-91. 8. Sepahdari AR, Politi LS, Aakalu VK, Kim HJ, Razek AA. Diffusion-weighted imaging of orbital masses: multi-institutional data support a 2-ADC threshold model to categorize lesions as benign, malignant, or indeterminate. AJNR American journal of neuroradiology. 2014;35(1):170-5. 9. Sepahdari AR, Aakalu VK, Setabutr P, Shiehmorteza M, Naheedy JH, Mafee MF. Indeterminate orbital masses: restricted diffusion at MR imaging with echo-planar diffusionweighted imaging predicts malignancy. Radiology. 2010;256(2):554-64. 10. Fatima Z, Ichikawa T, Ishigame K, Motosugi U, Waqar AB, Hori M, et al. Orbital masses: the usefulness of diffusion-weighted imaging in lesion categorization. Clinical neuroradiology. 2014;24(2):129-34. 11. Politi LS, Forghani R, Godi C, Resti AG, Ponzoni M, Bianchi S, et al. Ocular adnexal lymphoma: diffusion-weighted mr imaging for differential diagnosis and therapeutic monitoring. Radiology. 2010;256(2):565-74. 12. Zhang F, Sha Y, Qian J, Huang WH, Li XF, Wang SJ, et al. Role of magnetic resonance diffusion-weighted imaging in differentiating lacrimal masses. Journal of magnetic resonance imaging : JMRI. 2014;40(3):641-8. 13. Sepahdari AR, Aakalu VK, Kapur R, Michals EA, Saran N, French A, et al. MRI of orbital cellulitis and orbital abscess: the role of diffusion-weighted imaging. AJR American journal of roentgenology. 2009;193(3):W244-50. 14. Kapur R, Sepahdari AR, Mafee MF, Putterman AM, Aakalu V, Wendel LJ, et al. MR imaging of orbital inflammatory syndrome, orbital cellulitis, and orbital lymphoid lesions: the role of diffusion-weighted imaging. AJNR American journal of neuroradiology. 2009;30(1):64-70. 15. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. Journal of magnetic resonance imaging : JMRI. 1997;7(1):91-101. 16. O'Connor JP, Tofts PS, Miles KA, Parkes LM, Thompson G, Jackson A. Dynamic contrast-enhanced imaging techniques: CT and MRI. The British journal of radiology. 2011;84 Spec No 2:S112-20.
19 17. Yuan Y, Kuai XP, Chen XS, Tao XF. Assessment of dynamic contrast-enhanced magnetic resonance imaging in the differentiation of malignant from benign orbital masses. European journal of radiology. 2013;82(9):1506-11. 18. Yang BT, Wang YZ, Dong JY, Wang XY, Wang ZC. MRI study of solitary fibrous tumor in the orbit. AJR American journal of roentgenology. 2012;199(4):W506-11. 19. Lee FK, King AD, Ma BB, Yeung DK. Dynamic contrast enhancement magnetic resonance imaging (DCE-MRI) for differential diagnosis in head and neck cancers. European journal of radiology. 2012;81(4):784-8. 20. Accuracy of diagnosis of choroidal melanomas in the Collaborative Ocular Melanoma Study. COMS report no. 1. Archives of ophthalmology. 1990;108(9):1268-73. 21. Lichy MP, Aschoff P, Plathow C, Stemmer A, Horger W, Mueller-Horvat C, et al. Tumor detection by diffusion-weighted MRI and ADC-mapping--initial clinical experiences in comparison to PET-CT. Investigative radiology. 2007;42(9):605-13. 22. Kuhl CK, Mielcareck P, Klaschik S, Leutner C, Wardelmann E, Gieseke J, et al. Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions? Radiology. 1999;211(1):101-10. 23. Kiessling F, Jugold M, Woenne EC, Brix G. Non-invasive assessment of vessel morphology and function in tumors by magnetic resonance imaging. European radiology. 2007;17(8):2136-48.
20 FIGURE LEGENDS Fig 1. Selection process of patients and detailed results of the quantitative analysis.
Eligible patients n = 91
Excluded patients (n=4) <18 years old: 1 has undergone therapy between histopathology and MRI: 3
Index test = MRI n = 87
Excluded patients (n=22) No mass lesion seen on MRI: 9 lesion too small: 5 incomplete index test (no DCE): 4 no reference standard: 4
Final cohort n = 65
Malignant n = 33
Benign n = 32
Reference standard: histopathology: 15 definite clinical diagnosis: 18 (ADHM: 17, breast-Ca: 1)
Reference standard: histopathology: 18 lesion size unchanged over ≥ 6 months without and no clinical suspicion for disease treatment progression: 4 definite /pathognomonic MRI characteristics: 7 definite clinical diagnosis: 1 proven by surgery: 2
ADC <1.0 x 10-3mm²:25/33 >1.0 x 10-3mm²: 8/33
ADC <0.8 x 10-3mm²: 23/33 0.800-1.2 x 10-3mm²: 6/33 >1.2 x 10-3mm²: 4 /33
DCE
ADC <1.0 x 10-3mm²: 11/32 >1.0 x 10-3mm²: 21/32
DCE
ADC <0.8 x 10-3mm²: 4/32 0.8 x 10-3mm²-1.2 x 103mm²: 12/32 >1.2 x 10-3mm²: 16 /32
Curve-type I: 2/33 II: 6/33 III: 25/33
Perfusion Ktrans: 0,226 ± 0,136 Kep: 1,340 ± 0,804 Ve: 0,192 ± 0,074 iAUC: 11,144 ± 5,423
Curve-type I: 21/32 II: 7/32 III: 4/32
Perfusion Ktrans: 0,108 ± 0,084 Kep: 0,616 ± 0,558 Ve: 0,253 ± 0,137 iAUC: 6,712 ± 5,066
21 Fig 2. Qualitative analysis of both readers based on the Likert scale (1 to 5). Shaded: number of studies concerning standard anatomic images (1st reading session, n=130). Black: number of studies concerning standard anatomic images in combination with DW MR images (n=130). White: number of studies for standard anatomic images together with DCE MR images (n=130). Y-axis represents the number of studies. X-axis shows the rating scale (1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant).
22 Fig 3. A 45-year-old woman with orbital metastasis from breast-cancer imaged at 3 Tesla. (a) Axial T1-weighted image shows the metastasis isointense relative to muscle. (b) Axial T2weighted image shows the metastasis slightly hyperintense relative to muscle. (c) Axial DWI image shows bright signal intensity of the metastasis. (d) ADC map shows low signal intensity consistent with diffusion restriction within the metastasis (ADCROI = 0.734 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows signal-enhancement of the metastasis. (f) Time-intensity curve with a washout pattern (type 3).
23 Fig 4. A 40-year-old man with cavernous hemangioma imaged at 3 Tesla. (a) Axial T1weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor hyperintense relative to muscle. (c) Axial DWI image shows low signal intensity of the tumor. (d) ADC map shows tumor with high signal intensity consistent with absence of diffusion restriction within the hemangioma (ADCROI = 1.04 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows signal-enhancement of the tumor. (f) TimeIntensity curve with a persistent pattern (type 1).
24 Fig 5. A 64-year-old man with esthesioneuroblastoma imaged at 3 Tesla. (a) Axial T1weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor isointense relative to muscle. (c) Axial DWI image shows high signal intensity of the tumor. (d) ADC map shows tumor with low signal intensity consistent with diffusion restriction within the esthesioneuroblastoma (ADCROI = 0.971 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows marginal signal-enhancement of the tumor. (f) Time-Intensity curve with a washout pattern (type 3).
25 Fig 6. A 58-year-old man with neurofibroma imaged at 3 Tesla. (a) Axial T1-weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor hyperintense relative to muscle. (c) Axial DWI image shows inhomogeneous signal intensity of the tumor. (d) ADC map shows tumor with high signal intensity consistent with absence of diffusion restriction within the neurofibroma (ADCROI = 1.969 x 10-3 mm2/s). (e) Axial postcontrast-T1-weighted image shows subtle signal-enhancement of the tumor. (f) TimeIntensity curve with a persistent pattern (type 1).
26 Fig 7. A 59-year-old woman with dacryoadenitis imaged at 3 Tesla. (a) Axial T1-weighted image shows the inflamed gland isointense relative to muscle. (b) Axial T2-weighted image shows the inflamed gland hyperintense relative to muscle. (c) Axial DWI image shows a rather high signal intensity of the inflamed gland. (d) ADC map shows an intermediate to low signal of the inflamed gland (ADCROI = 0.874 x 10-3 mm2/s). (e) Axial post-contrast-T1weighted image shows a strong signal-enhancement of the inflamed gland. (f) Time-Intensity curve with a plateau pattern (type 2).
27 Fig 8. A 41-year-old woman with ocular melanoma imaged at 3 Tesla. (a) Axial T1-weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor with intermediate signal intensity. (c) Axial DWI image shows high signal intensity of the tumor. (d) ADC map shows the tumor with low signal intensity consistent with diffusion restriction within the ocular melanoma (ADCROI = 0.619 x 10-3 mm2/s). (e) Axial postcontrast-T1-weighted image shows strong signal-enhancement of the tumor. (f) TimeIntensity curve with a washout pattern (type 3).
28 Fig 9. The results of the ROC analysis of the reading sessions of reader 1 and 2 are displayed in this figure.
29 TABLE 1. MR sequence protocol Aquisition time Flip(min) angle (°)
Bandwith (Hz/px)
ETL
100
Parallel acquisition algorithm (acceleration factor) -
3.04
139
391
3
1
100
-
3.53
150
223
15
1
1
100
-
3.37
12
140
-
3
8
33
GRAPPA (4)
5.12
-
1028
128 (EPIFactor) -
Sequence
TR
TE
Inplane resolution (mm)
Field Number of view of (mm) slices
Slice thickness (mm)
NEX Phase oversampling (%)
T1-TSE
650
9.1
0.39 x 0.39
100 x 25 81.3
2
1
T2-TSE
3500 69
0.42 x 0.42
100 x 50 81.3
1
3D-T1GRE
15
4.23 0.42 x 0.42
80 x 48 65.04 144 144
DWI (SE- 5000 100 EPI)*
1.13 x 1.13
x 15
260 5.49 (75 12 measurements) = 4.56 sec temporal resolution TR, repetition time; TE, echo time; Inplane resolution = Field of view/matrix; NEX, number of excitations; px, pixel; ETL, echo train length; TSE, turbo spin echo; 3D indicates 3-dimensional; GRE, gradient echo; DWI, diffusion-weighted imaging; SE, spin echo; EPI, echo planar imaging; DCE, dynamic contrast enhanced; TWIST, time resolved angiography with stochastic trajectories. DCE (TWIST)
5.13
2.03 1.21 x 0.83
160 160
*b-values: 0 and 1000 s/mm² (3 directions)
x 20
2.5 (slice 1 resolution 62%)
20
GRAPPA (2)
30 TABLE 2. Summary of subjective assessments for each reader (n=65) Reader 1 sAI (1st reading)
1 1
2 11
3 24
4 26
5 3
Sensitivity 37.5%
Specificity 81.8%
sAI (2nd reading)
2
11
24
28
0
37.5%
78.8%
sAI + DWI
8
15
6
7
29
71.87%
96.96%
sAI + DCE
8
14
10
12
21
65.62%
93.93%
sAI + DWI + DCE
15
14
2
6
28
96.97%
90.63%
Reader 2 sAI (1st reading)
8
6
22
14
15
43.75%
75.75%
sAI (2nd reading)
4
10
27
18
6
37.5%
69.7%
sAI + DWI
9
13
5
11
27
68.75%
100%
sAI + DCE
7
17
13
11
21
71.87%
81.81%
sAI + DWI + DCE
16
14
1
4
30
100%
93.75%
1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant.
S0720-048X(15)30184-4 http://dx.doi.org/doi:10.1016/j.ejrad.2015.11.041 EURR 7331
To appear in:
European Journal of Radiology
Received date: Revised date: Accepted date:
4-8-2015 25-11-2015 30-11-2015
Please cite this article as: Ro Sa-Ra, Asbach Patrick, Siebert Eberhard, Bertelmann Eckart, Hamm Bernd, Erb-Eigner Katharina.Characterization of orbital masses by multiparametric MRI.European Journal of Radiology http://dx.doi.org/10.1016/j.ejrad.2015.11.041 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
Characterization of orbital masses by multiparametric MRI
Sa-Ra Roa, Patrick Asbach, MDa, Eberhard Siebert, MDb,
Eckart Bertelmann, MDc,
Bernd Hamm, MDa, Katharina Erb-Eigner, MDa
a
Department of Radiology, Charité - Universitätsmedizin Berlin, Campus Benjamin Franklin,
Hindenburgdamm 30, 12203 Berlin, Germany b
Department of Neuroradiology, Charité - Universitätsmedizin Berlin, Campus Charité Mitte,
Charitéplatz 1, 10117 Berlin, Germany c
Department of Ophthalmology, Charité - Universitätsmedizin Berlin, Campus Virchow-
Klinikum, Augustenburger Platz 1, 13353 Berlin, Germany
Corresponding author: Patrick Asbach, MD Department of Radiology Charité - Universitätsmedizin Berlin, Campus Benjamin Franklin Hindenburgdamm 30, 12203 Berlin, Germany Phone: +49 30 8445 3041 Fax: +49 30 450 7 527 953 E-mail address: [email protected]
2 Abstract Objectives DWI and dynamic contrast enhanced (DCE) MR imaging are techniques that allow insight to tumor vascularity and cellularity. We evaluated the diagnostic performance of multiparametric MRI (mp-MRI) in distinguishing benign from malignant orbital masses using standard anatomic imaging (sAI), DWI and DCE. Materials and Methods This prospective IRB approved study with written informed consent included 65 patients. mpMRI at 3 Tesla including DWI and DCE was performed in all patients. Parametric maps were generated for obtaining the perfusion parameters including Ktrans, kep, ve and iAUC and timesignal intensity curves were recorded to determine the curve pattern. Two radiologists rated the likelihood of malignancy on a five-point scale in three separate, randomized reading sessions (initially only sAI, afterwards sAI + either DWI or DCE and finally sAI + DWI + DCE). Data was statistically analyzed. Results 33 patients had malignant orbital masses and 32 patients had benign orbital masses (reference standard histopathology in 35 cases and clinical follow-up in 30 patients). The mean ADC of malignant masses differed significantly from the mean (SD) ADC of benign masses (0.825 [0.437] x 10-3mm²/s and 1.257 [0.576] × 10-3mm²/s, respectively) (P=0.001). Ktrans, kep and iAUC were significantly higher in malignant masses (P<0.01). The reading of sAI only resulted in a moderate specificity but poor sensitivity in differentiating benign from malignant lesions. Adding DWI and DCE images improved specificity and sensitivity considerably, being the highest for the combined reading of all sequences. Conclusion mp-MRI is a helpful tool in differentiating malignant orbital lesions from benign masses and should therefore be included in the routine diagnostic protocol for orbital imaging.
Keywords Orbital masses; Orbital MRI; Multiparametric MRI; Diffusion weighted imaging; Dynamic contrast enhanced imaging
3 Introduction
Orbital and intraocular masses are relatively uncommon compared to other mass lesions of the body. It has been described that two-thirds of orbital tumors are benign and one third is malignant (1). Characterization of orbital masses is crucial in the therapeutic strategy planning (2) as the patient management differs greatly depending on the dignity of the orbital mass (3). However, it is often difficult to differentiate malignant orbital masses from benign ones due to their similar clinical presentation with proptosis as the most common symptom (4). Magnetic resonance imaging (MRI) may help finding the diagnosis as there are some pathognomonic features for particular masses as for instance the cavernous hemangioma (5). But for cases without these pathognomonic characteristics it remains difficult to deliver a diagnosis based on MR imaging features because there are often unspecific and overlapping imaging findings (5). Aggravating this situation, rare tumor entities are unexpected and therefore may be misdiagnosed (5). It has even been proposed that none of the orbital imaging features including CT and MR imaging features had sufficient sensitivity to distinguish between malignant and benign orbital masses (6). Published data indicate that the use of advanced MRI sequence techniques like diffusion weighted imaging (DWI) with quantitative apparent diffusion coefficient (ADC) mapping and dynamic contrast-enhanced (DCE) MR imaging may provide additional information about the dignity and entity of orbital masses (5, 7-14). Based on the diffusivity of water in tissue DWI exploits the fact that the movement of water is normally restricted in malignant tumors, whereas DCE offers information on the rate of uptake and clearance of contrast media and can be analyzed to derive information on tissue and tumor vascularity. Tofts et al. defines a compartment model to describe abnormal capillary leakage and uses the following parameters: Ktrans = volume transfer constant between blood plasma and extravascular extracellular space (EES) in abnormal tissue, kep = efflux rate constant from EES to blood
4 plasma, ve = EES volume per unit volume of tissue and iAUC = initial area under the gadolinium concentration curve during the first 60 seconds (15).. Taking advantage of either of these, DWI and DCE, there are several research groups who have investigated multiparametric magnetic resonance imaging (mp-MRI) of the orbital cavity by adding one functional parameter such as DWI or DCE to the standard morphological sequences (4, 5, 7, 17, 18). There were suggestions that DCE or DWI may produce complementary information (19). However, to our knowledge, the diagnostic performance of combining conventional magnetic resonance imaging with DWI and DCE has not been investigated. The purpose of our study was to prospectively evaluate the diagnostic performance of mpMRI in distinguishing benign from malignant orbital masses using DWI and DCE MR imaging.
5 Material and Methods
Patients The institutional review board approved this prospective single-institution clinical study. Written informed consent was obtained from all patients. Between May 2013 and July 2014 a total of 91 patients with known or suspected orbital space-occupying lesion were recruited in the outpatient ophthalmology clinic. This study was registered at clinicaltrials.gov under ClinicalTrials.gov-ID (blinded for review). A senior ophthalmologist who is specialized on orbital tumors examined all patients before the MRI examination. Patients were excluded if they were not at least 18 years old (n=1) or if they had undergone therapy between histopathology and MRI examination (n=3). Other exclusion criteria were contraindications to MRI or contrast agent applications (GFR<30 ml/min; allergies), pregnancy and claustrophobia. Ultimately 87 patients were enrolled in this study and underwent MR imaging with standard anatomical imaging (sAI) including T1-weighted, T2-weighted and post-contrast T1-weighted sequences and multiparametric MR imaging (mpMRI) including diffusion weighted imaging (DWI) with ADC mapping and dynamic contrast enhanced (DCE) imaging. In addition to the previously mentioned criteria, nine patients were excluded because there was no mass lesion seen on MRI, five were excluded because the lesion was too small (≤ 4 mm) and partial volume averaging effects hinder accurate analysis on mp-MRI, another four patients had an incomplete index test and four patients had no reference standard, resulting in a final cohort with 65 patients. Final diagnosis was established through histopathological results from surgery (n=2) or biopsy (n=33), pathognomonic imaging findings (n=7) or clinical follow up over a period of a minimum of 6 months (benign lesions with no change, n=4), ophthalmoscopy (uveal
6 melanoma, n=17) or unequivocal clinical diagnosis (multi-organ metastatic breast cancer, n=1; pseudotumor orbitae, n=1).
MR imaging technique Imaging was performed on a 3 Tesla scanner (MAGNETOM Skyra; Siemens Healthcare, Erlangen, Germany) with either a single-channel 7-cm loop surface coil (for sAI) or a 20channel head coil (for DWI and DCE). The respective sequence parameters of the MR protocol are summarized in Table 1. Contrast medium was administered intravenously body weight adapted (0.1 mmol gadoteric acid/kg body weight) after the 3rd DCE measurement (approximately 13 seconds) as bolus injection in an antecubital vein with an injection speed of 2ml/s using a 0.60x25mm tube.
Image analysis Quantitative Analysis The sAI were used to measure the diameter of the lesion in three dimensions. The mean diameter was calculated. In the DW images a freehand drawn ROI was placed in the lesion at b = 0 and 1000 sec/mm², while the edges of each lesion were avoided to minimize the effect of partial volume averaging. Similarly, an identical ROI was placed within the vitreous humor. The number of pixels, the size of the ROI and the SI at b = 0 and 1000 sec/mm² were recorded for each patient. The ADC values of the respective ROIs were calculated according to the formula
ADC=
.
A 2-ADC threshold model was used with 0.8 x 10-3mm²/s and 1.2 x 10-3mm²/s as thresholds to categorize lesions in three groups.
7 For the image analysis of the DCE MR images we used a dedicated postprocessing software program (Tissue 4D; Siemens Healthcare) which supplies pharmacokinetic calculation on a pixel-by-pixel basis using a 2-compartment model based on the Tofts model. Concerning the arterial input function it was decided to choose the intermediate sampling method. A freehand ROI was drawn over the area of maximum DCE MR imaging signal within each lesion and voxelwise perfusion maps containing the following parameters were generated: mean Ktrans, kep, ve and iAUC. Using the same postprocessing software program we recorded time-signal intensity curves for each lesion in addition to the arrival time, the time of the peak and the last time point of the contrast agent and the respective signal intensities. There were three different types of curve pattern. Type 1 curves showed a persistent pattern (i.e., SIlast time = SIpeak time or SIlast >= SIpeak), type 2 curves showed a plateau pattern (i.e., SIpeak time-SIlast time ≤10% of SIpeak time) and type 3 curves showed a washout pattern (i.e., SIpeak time-SIlast time>10% of SIpeak time). T1-mapping was included in the protocol but we refrained from ROI analysis since no standard of reference parameter have been established yet.
Qualitative Analysis Two radiologists with 11 and 8 years of experience in MR imaging who were blinded to the diagnosis, clinical history and histopathologic findings independently reviewed the MR images in randomized order in three reading sessions, with a three week interval between the first two sessions and a six months interval between the second and third reading session. Prior to the first reading the readers underwent a short training session that displayed five benign and five malignant orbital lesions in the respective sequences. In the first session each reader reviewed the sAI sequences (T1-weighted, T2-weighted and T1 post-contrast) only, followed by the addition of DW images and the corresponding ADC maps. For the 2nd session the readers reviewed initially the sAI sequences again followed by the DCE MR
8 images. For the 3rd session the readers reviewed the sAI sequences together with the DWI and DCE MR images. At each reading the readers were instructed to score the images for malignancy on a 1 to 5 Likert scale to convert subjective assessments into quantifiable parameters (1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant). They also independently graded the sequence quality of the standard anatomical images and the mp-MR images (DCE and DWI) with regard to artifacts on a scale of 1 to 5 (‘The study shows a perfect representation of the orbit without artifacts’, 1 = strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree and 5 = strongly agree).
Statistical analysis Descriptive statistics for the outcome of the qualitative analysis compared to the reference standard were performed for each reader and each sequence and the corresponding sensitivity and specificity were determined. Differences between the sensitivities and specificities of each viewing setting were tested for significance based on the binomial test. The nonparametric Mann-Whitney-U test for unpaired variables was applied to determine whether the diameter, ve, ADC, Ktrans, kep and iAUC of benign lesions differed significantly from those of malignant lesions using the Statistical Package for the Social Sciences software (SPSS version 11.5). Using the Pearson Chi² Test the curve pattern was tested for significance. The software 'R' (R version 3.1.2) was used for comparison of ROC curves and to test inter-reader and intra-reader variability (Cohen’s kappa test). Since there were multiple comparisons Bonferroni adjustment of P values was used. differences.
P < 0.01 indicated statistically significant
9 Results
Patients Out of 91 eligible patients 87 patients underwent MR imaging, and 22 were excluded (Fig 1). The final cohort included 65 patients. In 32 cases (32/65, 49%) malignancy was confirmed, of whom there were 17 men (age range, 26-80 years; mean age, 60 years) and 15 women (age range, 38-68 years; mean age, 55 years). The other 33 patients (33/65, 51%) had benign lesions, of whom there were 20 men (age range, 20-75 years; mean age, 53 years) and 13 women (age range, 29-88 years; mean age, 61 years). The 33 malignant lesions included the following diagnoses: conjunctival carcinoma (n=1), squamous epithelial carcinoma (n=1), adenoid cystic carcinoma (n=1), not otherwise specified carcinoma but histopathology states it as malignant (n=1), follicular lymphoma (n=2), marginal zone lymphoma (n=4), mantle cell lymphoma (n=2), esthesioneuroblastoma (n=1), uveal melanoma (n=17), melanoma (n=1), metastasis of breast cancer (n=2) of whom one was proven by histopathology while the other one was confirmed by follow up (within three month since the last MR examination a newly developed breast cancer was found and the patient developed multi organ metastatic disease).
Except for uveal melanoma where
fundoscopy is the gold standard for diagnosis (20) and the above-mentioned patient who developed metastasis from breast cancer all other malignant lesions were proven by histopathology. The diagnoses of the 32 benign lesions were as follows: hemangioma (n=5), meningioma (n=3), angiomatosis (n=1), neurofibroma (n=1), lipoma (n=2), lymphatic tissue (n=2), lymphangioma (n=1), sarcoidosis (n=1), endocrine orbitopathy (n=1), inflammation of lacrimal gland (n=3), amyloidosis (n=1), hematoma (n=1), pseudotumor (n=2), mucocele (n=1), cyst (n=1), schwannoma (n=1), orbital vein thrombosis (n=1), granulation tissue (n=1), myxoma (n=1), pyocele (n=1), not otherwise specified lesion but histopathology claims it as
10 benign (n=1). There were pathognomonic imaging findings for four hemangiomas, two optic nerve sheath meningiomas and the orbital vein thrombosis. The pyocele and one pseudotumor were proven by surgery. The other pseudotumor was determined as definite clinical diagnosis. The angiomatosis, lymphangioma, endocrine orbitopathy and hematoma were confirmed by follow-up. The remaining 18 cases were proven by histopathology.
Quantitative analysis The mean diameter (SD) of benign lesions was 1.74 (0.86) cm (range 0.47 – 5.19 cm) and was 1.64 (0.91) cm (range 0.72 – 4.59 cm) of malignant lesions. The difference was not statistically significant (p=0.259). The mean (SD) number of pixel per ROI in the DW images was 29 (17). The mean (SD) size of a ROI was 29 (20) mm². The mean (SD) ADC of the vitreous humor was 2.686 (0.257) x 10-3mm²/s. There were significant differences regarding the ADC between malignant and benign lesions (p=0.001). In the malignant lesions the mean (SD) ADC was 0.825 (0.437) x 10-3mm²/s, while the mean (SD) ADC of benign lesions was 1.258 (0.577) x 10-3mm²/s. Applying an ADC value of 1.0 x 10-3mm²/s as a threshold for differentiation between malignant and benign lesions resulted in 75% of the malignant studies to be found below the threshold and 25% above the threshold. The mean (SD) ADC of the ocular melanoma subgroup (n=17 of 33 malignant lesions) was 0.771 (0.016) x 10-3mm²/s. ROC analysis revealed that a ADC of less than 8.141 x 10-3mm²/s was more than 88% specific for malignancy, and a ADC of less than 1.151 x 10-3mm²/s was more than 88% sensitive for malignancy (AUC=0.767). For a more accurate approach a 2-ADC threshold model was used with 0.8 x 10-3mm²/s and 1.2 x 10-3mm²/s as thresholds. As a result only 12% of the malignant lesions had an ADC above 1.2 x 10-3mm²/s (Fig 1). For the benign lesions the same thresholds
11 were used as for the malignant lesions. Again the 2-ADC threshold model reduced the number of benign studies, which wrongly indicated malignancy, from 34% to 12.5%. The exact numbers for the DCE perfusion parameters are given in Figure 1. In 3 studies the patients underwent 2D Perfusion imaging and were therefore excluded from the statistical evaluation of the perfusion parameters. In the remaining 62 studies the perfusion parameters Ktrans, kep and iAUC were significantly higher in malignant lesions (p<0.01). ROC analysis suggested that a Ktrans of more than 0.257 was more than 90% specific for malignancy, and a Ktrans of more than 0.077 was more than 90% sensitive for malignancy (AUC=0.781). A kep of more than 1.206 was more than 90% specific for malignancy, and a kep of more than 0.617 was more than 90% sensitive for malignancy (AUC=0.846). Also, an iAUC of more than 14.468 was more than 90% specific for malignancy, and an iAUC of more than 4.635 was more than 90% sensitive for malignancy (AUC=0.735). On the contrary the mean ve in malignant lesions was lower than in benign lesions and there were no significant differences for ve (p=0.108). The curve pattern could be obtained for all of the 65 patients. The curve pattern for malignant lesions differed significantly from benign lesions (p<0.01). Malignant lesions mainly (25/33) showed a washout pattern and benign lesions mostly (21/32) a persistent pattern.
Qualitative analysis Since the sAI images were scored twice in two independent sessions there were a total of 130 sAI readings. While for the combination of the standard anatomic images with DW images and with DCE MR images there were 65 studies each. As shown in Figure 2 both readers rated the images mostly as indeterminate or as probably malignant if they had only the standard anatomic images at their disposal. The combination of the standard anatomic images and DW images proved to be most conclusive to decide whether malignancy was detected on the images. While the addition of DCE MR images to the standard anatomic images was
12 inferior compared to DWI for determining malignancy, the combination of both, adding DW and DCE images to sAI produced the best results to differentiate between malignant and benign lesions. The precise details of the ratings and the respective sensitivity and specificity for each reader and each image-viewing-setting are given in table 2. Figure 3-7 gives examples of the sAI images, the DWI, the ADC-map, the post-contrast T1-weighted images and the time-intensity curve of benign and malignant lesions. For reader 1 the standard anatomic images in combination with the DW images yielded the highest sensitivity (71.87%) and the highest specificity (96.96%). For reader 2 the sensitivity was highest with the combination of standard anatomic images and DCE images (71.87%), while the highest specificity was found with the standard anatomic images in addition to the DW images (100%). All in all both the addition of DW MR images for diagnostic purposes and the addition of both, the DW and DCE MR images produced statistically significant improvement regarding sensitivity and specificity for both readers (P<0.01). ROC analyses of the readings of the two readers are displayed in Figure 9. Comparing ROC curves the addition of DW MR images to sAI images improved the AUC significantly (Reader 1: p=0.015; Reader2: p=0.018). Also, the AUC results of the reading using sAI, DW and DCE MR images simultaneously differed significantly from the AUC results using sAI images only (Reader 1: p=0.003; Reader 2: p=0.006). The differences between the first and second sAI readings of the first reader, as well as the second reader indicated a different level of confidence (intra-observer variability Reader 1: κ=0.227 ‘fair agreement’; Reader 2: κ=0.582 ‘moderate agreement’). The inter-observer variability between Reader 1 and Reader 2 decreased by the addition of DCE and DWI images (first sAI reading: κ=0.292 ‚fair agreement‘; second sAI reading: κ=0.494‚ moderate agreement‘; sAI+DWI reading: κ=0.534 ‚moderate agreement‘; sAI+DCE reading: κ=0.627 ‚substantial agreement‘; sAI+DWI+DCE reading: κ=0.644 ‚substantial agreement‘).
13 Both readers rated that the standard anatomic images and the DCE MR images showed a perfect delineation of the structures of the orbital cavity in 64 of the 65 studies (score of 4 or 5). In one case they rated the sequence quality with regard to artifacts as 3. Reasonable diagnostic image quality with DW imaging was obtained in 58 of the 65 studies for reader 1. In the remaining studies reader 1 rated 3 in 5 studies and 2 in 2 studies. Reader 2 rated in 51 studies 4 or 5, in 9 studies 3, in 4 studies 2 and in one case 1.
14 Discussion
Aim of the study was to demonstrate significant quantifiable differences between malignant and benign orbital masses using mp-MRI: The lower the ADC value of an orbital mass, the more likely the mass was malignant, while benign orbital masses showed significantly higher ADC values. Supplementary, high values of Ktrans, kep and iAUC revealed a high probability for malignancy. Investigating the ve values among the orbital masses a tendency was observed: Benign masses demonstrated higher values while malignant lesions more often presented lower values. Furthermore our results propose that DWI and DCE MR imaging are complimentary: For instance one patient with sarcoid exhibited an ADC value that could be interpreted as either benign or malignant which may be due to the fact that the dominant reaction is a dense fibrous tissue with restricted diffusion. However when investigated with DCE MR imaging the lesion had a persistent curve pattern and low values for Ktrans, kep and iAUC. This way the lesion could be rightly categorized as benign. Similar to Yuan et al. most of the malignant lesions showed a washout pattern (curve type III; 25 of 33) and most of the benign lesions showed a persistent pattern (curve type I; 21 of 32) (17). DWI visualizes differences in tissue cellularity and has been used throughout the body to categorize lesions in 'benign' or 'malignant' (21). According to their degree of cellularity masses can be distinguished in malignant and benign. High cellularity of a tissue causes high restriction of water molecules in this tissue, which in turn is associated with lower ADC values. Sepahdari et al. described that with the addition of DWI, improved accuracy of MR imaging in the characterization of indeterminate orbital lesions may help triage patients to either early intervention or initial conservative treatments (9). In our study using a 2-ADCtreshold model had better results than using only one threshold for distinguishing between malignant and benign masses (for example: malignant masses which simulate benign masses:
15 1 ADC- Threshold: 24.24%, 2-ADC-Thresholds: 12.12%). This observation corresponds to other findings of Sepahdari et al. who described a 2-ADC threshold model to categorize orbital masses with high confidence (8). Fatima et al. reported a mean ADC value of 0.77 x 10-3mm²/s for malignant lesions and a mean ADC for benign lesions 1.23 x 10-3mm²/s (10). Sepahdari et al. reported a mean ADC value of 1.43 ± 0.41 x 10-3mm²/s for malignant lesions and a mean ADC for benign lesions of 0.90 ± 0.36 x 10-3mm²/s (8). In our study, the mean (SD) ADC was 0.825 (0.437) x 10-3mm²/s for malignant lesions, while the mean (SD) ADC of benign lesions was 1.258 (0.577) x 10-3mm²/s. ROC analysis revealed that an ADC of less than 8.141 x 10-3mm²/s was more than 88% specific (and less than 6.456 x 10-3mm²/s was more than 90% specific) for malignancy, and a ADC of less than 1.151 x 10-3mm²/s was more than 88% sensitive (and less than 1.462 x 10-3mm²/s was more than 90% sensitive) for malignancy. It was noticeable that with only one cut- off value, the rate of clearly erroneous categorized lesions was significantly higher than with 2 cut-off values. For clinical practice, this means, that 8 malignant lesions would have been categorized as benign with only one cutoff value and 4 malignant lesions with 2 cut-off values, which would have a more hesitant therapy approach as a result. The introduction of a 2-ADC threshold model would mean that in the 6 borderline cases further diagnostic work-up would be used to determine the dignity of the lesion and the probability is increased to inflict upon the correct treatment of each lesion. Another noninvasive advanced MRI sequence is the DCE MR imaging. DCE MRI visualizes tumor vascularity pattern and has been used to characterize various tumors throughout the body (22). Our study showed that analysis of the curve type (persistent, plateau and washout pattern) allowed a confident characterization of the respective lesion into 'benign' or 'malignant'. There were only 2 malignant lesions that were classified as benign, whilst with DWI 8 (ADC>1.0 x 10-3mm²/s) or 4 malignant lesions (ADC>1.2 x 10-3mm²/s) were classified wrongly as benign. The analysis of objective DCE parameters (Ktrans, kep and iAUC)
16 was significantly higher in malignant lesions than in benign ones. Higher vascular permeability and higher perfusion is typically observed in malignant lesions (23). The results of the qualitative analysis suggest that the visual impression of DWI and DCE images increases the confidence of the reader. The reading of native images only resulted in a moderate specificity and an unsatisfying sensitivity. Adding both, DWI and DCE images improved specificity and sensitivity considerably, where the DWI showed better results increasing confidence of both readers. There are already studies that have dealt with either DWI or DCE, but to our knowledge there is no study that has tested both imaging techniques on one group of patients. Therefore, selection bias that may have been occurred in studies that dealt with DWI or DCE only could be avoided in our study. Furthermore, our study considered qualitative image analysis as well, whilst many studies are restricting on quantitative analysis only. Our study has several limitations. A patient population bias cannot be excluded which is reflected in the distribution of the orbital masses regarding their dignity. The tumors included reflect the referral pattern to our orbital clinic, which included equally benign and malignant lesions, and not 2 third to 1 third. Another limitation may be the manual delineation of lesions and calculation of DCE parameters over the entire tumor. The SI of the ROI is averaged and peak enhancement value is influenced by the complexity and heterogeneity (e.g. necrosis) of the mass. Furthermore, no multi-channel small loop surface coil was available, although the DWI and DCE sequences were acquired with the multi-channel head coil covering both orbits. In conclusion, this prospective study confirmed that mp-MRI through adding DWI or both DWI and DCE MR imaging improved not only readers’ confidence in categorizing orbital mass lesions (qualitative analysis) but also showed a good diagnostic performance in terms of quantifiable results (quantitative analysis). DWI and DCE MR imaging are helpful tools in differentiating malignant orbital lesions from benign masses. We therefore recommend including DWI and DCE MR imaging in the routine diagnostic approach of orbital masses.
17 Conflict of interest The authors have no conflict of interest to disclose.
AUTHOR DECLARATION We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing we confirm that we have followed the regulations of our institutions concerning intellectual property. We further confirm that any aspect of the work covered in this manuscript that has involved human patients has been conducted with the ethical approval of all relevant bodies and that such approvals are acknowledged within the manuscript. We understand that the Corresponding Author is the sole contact for the Editorial process (including Editorial Manager and direct communications with the office). He is responsible for communicating with the other authors about progress, submissions of revisions and final approval of proofs. Patrick Asbach, MD (Corresponding Author) [email protected] Signed by all authors as follows:
Sa-Ra Ro Patrick Asbach Eberhard Siebert Eckart Bertelmann Bernd Hamm Katharina Erb-Eigner 31/07/2015
Acknowledgments K. Erb-Eigner is supported by the ‘Rahel Hirsch Program’ funded by the Charité – Universitätsmedizin Berlin.
18 References
1. Shields JA, Shields CL, Scartozzi R. Survey of 1264 patients with orbital tumors and simulating lesions: The 2002 Montgomery Lecture, part 1. Ophthalmology. 2004;111(5):9971008. 2. Lemke AJ, Kazi I, Felix R. Magnetic resonance imaging of orbital tumors. European radiology. 2006;16(10):2207-19. 3. Valvassori GE, Sabnis SS, Mafee RF, Brown MS, Putterman A. Imaging of orbital lymphoproliferative disorders. Radiologic clinics of North America. 1999;37(1):135-50, x-xi. 4. Razek AA, Elkhamary S, Mousa A. Differentiation between benign and malignant orbital tumors at 3-T diffusion MR-imaging. Neuroradiology. 2011;53(7):517-22. 5. Xian J, Zhang Z, Wang Z, Li J, Yang B, Chen Q, et al. Evaluation of MR imaging findings differentiating cavernous haemangiomas from schwannomas in the orbit. European radiology. 2010;20(9):2221-8. 6. Ben Simon GJ, Annunziata CC, Fink J, Villablanca P, McCann JD, Goldberg RA. Rethinking orbital imaging establishing guidelines for interpreting orbital imaging studies and evaluating their predictive value in patients with orbital tumors. Ophthalmology. 2005;112(12):2196-207. 7. Roshdy N, Shahin M, Kishk H, Ghanem AA, El-Khouly S, Mousa A, et al. MRI in diagnosis of orbital masses. Current eye research. 2010;35(11):986-91. 8. Sepahdari AR, Politi LS, Aakalu VK, Kim HJ, Razek AA. Diffusion-weighted imaging of orbital masses: multi-institutional data support a 2-ADC threshold model to categorize lesions as benign, malignant, or indeterminate. AJNR American journal of neuroradiology. 2014;35(1):170-5. 9. Sepahdari AR, Aakalu VK, Setabutr P, Shiehmorteza M, Naheedy JH, Mafee MF. Indeterminate orbital masses: restricted diffusion at MR imaging with echo-planar diffusionweighted imaging predicts malignancy. Radiology. 2010;256(2):554-64. 10. Fatima Z, Ichikawa T, Ishigame K, Motosugi U, Waqar AB, Hori M, et al. Orbital masses: the usefulness of diffusion-weighted imaging in lesion categorization. Clinical neuroradiology. 2014;24(2):129-34. 11. Politi LS, Forghani R, Godi C, Resti AG, Ponzoni M, Bianchi S, et al. Ocular adnexal lymphoma: diffusion-weighted mr imaging for differential diagnosis and therapeutic monitoring. Radiology. 2010;256(2):565-74. 12. Zhang F, Sha Y, Qian J, Huang WH, Li XF, Wang SJ, et al. Role of magnetic resonance diffusion-weighted imaging in differentiating lacrimal masses. Journal of magnetic resonance imaging : JMRI. 2014;40(3):641-8. 13. Sepahdari AR, Aakalu VK, Kapur R, Michals EA, Saran N, French A, et al. MRI of orbital cellulitis and orbital abscess: the role of diffusion-weighted imaging. AJR American journal of roentgenology. 2009;193(3):W244-50. 14. Kapur R, Sepahdari AR, Mafee MF, Putterman AM, Aakalu V, Wendel LJ, et al. MR imaging of orbital inflammatory syndrome, orbital cellulitis, and orbital lymphoid lesions: the role of diffusion-weighted imaging. AJNR American journal of neuroradiology. 2009;30(1):64-70. 15. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. Journal of magnetic resonance imaging : JMRI. 1997;7(1):91-101. 16. O'Connor JP, Tofts PS, Miles KA, Parkes LM, Thompson G, Jackson A. Dynamic contrast-enhanced imaging techniques: CT and MRI. The British journal of radiology. 2011;84 Spec No 2:S112-20.
19 17. Yuan Y, Kuai XP, Chen XS, Tao XF. Assessment of dynamic contrast-enhanced magnetic resonance imaging in the differentiation of malignant from benign orbital masses. European journal of radiology. 2013;82(9):1506-11. 18. Yang BT, Wang YZ, Dong JY, Wang XY, Wang ZC. MRI study of solitary fibrous tumor in the orbit. AJR American journal of roentgenology. 2012;199(4):W506-11. 19. Lee FK, King AD, Ma BB, Yeung DK. Dynamic contrast enhancement magnetic resonance imaging (DCE-MRI) for differential diagnosis in head and neck cancers. European journal of radiology. 2012;81(4):784-8. 20. Accuracy of diagnosis of choroidal melanomas in the Collaborative Ocular Melanoma Study. COMS report no. 1. Archives of ophthalmology. 1990;108(9):1268-73. 21. Lichy MP, Aschoff P, Plathow C, Stemmer A, Horger W, Mueller-Horvat C, et al. Tumor detection by diffusion-weighted MRI and ADC-mapping--initial clinical experiences in comparison to PET-CT. Investigative radiology. 2007;42(9):605-13. 22. Kuhl CK, Mielcareck P, Klaschik S, Leutner C, Wardelmann E, Gieseke J, et al. Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions? Radiology. 1999;211(1):101-10. 23. Kiessling F, Jugold M, Woenne EC, Brix G. Non-invasive assessment of vessel morphology and function in tumors by magnetic resonance imaging. European radiology. 2007;17(8):2136-48.
20 FIGURE LEGENDS Fig 1. Selection process of patients and detailed results of the quantitative analysis.
Eligible patients n = 91
Excluded patients (n=4) <18 years old: 1 has undergone therapy between histopathology and MRI: 3
Index test = MRI n = 87
Excluded patients (n=22) No mass lesion seen on MRI: 9 lesion too small: 5 incomplete index test (no DCE): 4 no reference standard: 4
Final cohort n = 65
Malignant n = 33
Benign n = 32
Reference standard: histopathology: 15 definite clinical diagnosis: 18 (ADHM: 17, breast-Ca: 1)
Reference standard: histopathology: 18 lesion size unchanged over ≥ 6 months without and no clinical suspicion for disease treatment progression: 4 definite /pathognomonic MRI characteristics: 7 definite clinical diagnosis: 1 proven by surgery: 2
ADC <1.0 x 10-3mm²:25/33 >1.0 x 10-3mm²: 8/33
ADC <0.8 x 10-3mm²: 23/33 0.800-1.2 x 10-3mm²: 6/33 >1.2 x 10-3mm²: 4 /33
DCE
ADC <1.0 x 10-3mm²: 11/32 >1.0 x 10-3mm²: 21/32
DCE
ADC <0.8 x 10-3mm²: 4/32 0.8 x 10-3mm²-1.2 x 103mm²: 12/32 >1.2 x 10-3mm²: 16 /32
Curve-type I: 2/33 II: 6/33 III: 25/33
Perfusion Ktrans: 0,226 ± 0,136 Kep: 1,340 ± 0,804 Ve: 0,192 ± 0,074 iAUC: 11,144 ± 5,423
Curve-type I: 21/32 II: 7/32 III: 4/32
Perfusion Ktrans: 0,108 ± 0,084 Kep: 0,616 ± 0,558 Ve: 0,253 ± 0,137 iAUC: 6,712 ± 5,066
21 Fig 2. Qualitative analysis of both readers based on the Likert scale (1 to 5). Shaded: number of studies concerning standard anatomic images (1st reading session, n=130). Black: number of studies concerning standard anatomic images in combination with DW MR images (n=130). White: number of studies for standard anatomic images together with DCE MR images (n=130). Y-axis represents the number of studies. X-axis shows the rating scale (1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant).
22 Fig 3. A 45-year-old woman with orbital metastasis from breast-cancer imaged at 3 Tesla. (a) Axial T1-weighted image shows the metastasis isointense relative to muscle. (b) Axial T2weighted image shows the metastasis slightly hyperintense relative to muscle. (c) Axial DWI image shows bright signal intensity of the metastasis. (d) ADC map shows low signal intensity consistent with diffusion restriction within the metastasis (ADCROI = 0.734 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows signal-enhancement of the metastasis. (f) Time-intensity curve with a washout pattern (type 3).
23 Fig 4. A 40-year-old man with cavernous hemangioma imaged at 3 Tesla. (a) Axial T1weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor hyperintense relative to muscle. (c) Axial DWI image shows low signal intensity of the tumor. (d) ADC map shows tumor with high signal intensity consistent with absence of diffusion restriction within the hemangioma (ADCROI = 1.04 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows signal-enhancement of the tumor. (f) TimeIntensity curve with a persistent pattern (type 1).
24 Fig 5. A 64-year-old man with esthesioneuroblastoma imaged at 3 Tesla. (a) Axial T1weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor isointense relative to muscle. (c) Axial DWI image shows high signal intensity of the tumor. (d) ADC map shows tumor with low signal intensity consistent with diffusion restriction within the esthesioneuroblastoma (ADCROI = 0.971 x 10-3 mm2/s). (e) Axial post-contrast-T1-weighted image shows marginal signal-enhancement of the tumor. (f) Time-Intensity curve with a washout pattern (type 3).
25 Fig 6. A 58-year-old man with neurofibroma imaged at 3 Tesla. (a) Axial T1-weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor hyperintense relative to muscle. (c) Axial DWI image shows inhomogeneous signal intensity of the tumor. (d) ADC map shows tumor with high signal intensity consistent with absence of diffusion restriction within the neurofibroma (ADCROI = 1.969 x 10-3 mm2/s). (e) Axial postcontrast-T1-weighted image shows subtle signal-enhancement of the tumor. (f) TimeIntensity curve with a persistent pattern (type 1).
26 Fig 7. A 59-year-old woman with dacryoadenitis imaged at 3 Tesla. (a) Axial T1-weighted image shows the inflamed gland isointense relative to muscle. (b) Axial T2-weighted image shows the inflamed gland hyperintense relative to muscle. (c) Axial DWI image shows a rather high signal intensity of the inflamed gland. (d) ADC map shows an intermediate to low signal of the inflamed gland (ADCROI = 0.874 x 10-3 mm2/s). (e) Axial post-contrast-T1weighted image shows a strong signal-enhancement of the inflamed gland. (f) Time-Intensity curve with a plateau pattern (type 2).
27 Fig 8. A 41-year-old woman with ocular melanoma imaged at 3 Tesla. (a) Axial T1-weighted image shows the tumor isointense relative to muscle. (b) Axial T2-weighted image shows the tumor with intermediate signal intensity. (c) Axial DWI image shows high signal intensity of the tumor. (d) ADC map shows the tumor with low signal intensity consistent with diffusion restriction within the ocular melanoma (ADCROI = 0.619 x 10-3 mm2/s). (e) Axial postcontrast-T1-weighted image shows strong signal-enhancement of the tumor. (f) TimeIntensity curve with a washout pattern (type 3).
28 Fig 9. The results of the ROC analysis of the reading sessions of reader 1 and 2 are displayed in this figure.
29 TABLE 1. MR sequence protocol Aquisition time Flip(min) angle (°)
Bandwith (Hz/px)
ETL
100
Parallel acquisition algorithm (acceleration factor) -
3.04
139
391
3
1
100
-
3.53
150
223
15
1
1
100
-
3.37
12
140
-
3
8
33
GRAPPA (4)
5.12
-
1028
128 (EPIFactor) -
Sequence
TR
TE
Inplane resolution (mm)
Field Number of view of (mm) slices
Slice thickness (mm)
NEX Phase oversampling (%)
T1-TSE
650
9.1
0.39 x 0.39
100 x 25 81.3
2
1
T2-TSE
3500 69
0.42 x 0.42
100 x 50 81.3
1
3D-T1GRE
15
4.23 0.42 x 0.42
80 x 48 65.04 144 144
DWI (SE- 5000 100 EPI)*
1.13 x 1.13
x 15
260 5.49 (75 12 measurements) = 4.56 sec temporal resolution TR, repetition time; TE, echo time; Inplane resolution = Field of view/matrix; NEX, number of excitations; px, pixel; ETL, echo train length; TSE, turbo spin echo; 3D indicates 3-dimensional; GRE, gradient echo; DWI, diffusion-weighted imaging; SE, spin echo; EPI, echo planar imaging; DCE, dynamic contrast enhanced; TWIST, time resolved angiography with stochastic trajectories. DCE (TWIST)
5.13
2.03 1.21 x 0.83
160 160
*b-values: 0 and 1000 s/mm² (3 directions)
x 20
2.5 (slice 1 resolution 62%)
20
GRAPPA (2)
30 TABLE 2. Summary of subjective assessments for each reader (n=65) Reader 1 sAI (1st reading)
1 1
2 11
3 24
4 26
5 3
Sensitivity 37.5%
Specificity 81.8%
sAI (2nd reading)
2
11
24
28
0
37.5%
78.8%
sAI + DWI
8
15
6
7
29
71.87%
96.96%
sAI + DCE
8
14
10
12
21
65.62%
93.93%
sAI + DWI + DCE
15
14
2
6
28
96.97%
90.63%
Reader 2 sAI (1st reading)
8
6
22
14
15
43.75%
75.75%
sAI (2nd reading)
4
10
27
18
6
37.5%
69.7%
sAI + DWI
9
13
5
11
27
68.75%
100%
sAI + DCE
7
17
13
11
21
71.87%
81.81%
sAI + DWI + DCE
16
14
1
4
30
100%
93.75%
1 = definitely benign, 2 = probably benign, 3 = indeterminate, 4 = probably malignant and 5 = definitely malignant.