Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

Diagnostic and Interventional Imaging (2018) xxx, xxx—xxx

ORIGINAL ARTICLE /Abdominal imaging

Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients A. Larbi ∗, C. Orliac, J. Frandon, F. Pereira, A. Ruyer, J. Goupil , F. Macri , J.P. Beregi , J. Greffier Department of radiology, CHU de Nîmes, EA2415, MIG Nimes, place du Professeur-Robert-Debré, 30900 Nîmes, France

KEYWORDS Oncology; Focal liver lesion; Computed tomography (CT); Iterative reconstruction; Lesion detection

Abstract Purpose: The purpose of this study was to evaluate and compare the diagnostic accuracy of ultra-low dose (ULD) computed tomography (CT) with that of standard dose (STD) CT in the detection and characterization of focal liver lesions in neoplastic patients. Materials and methods: A total of 177 neoplastic patients who underwent two abdominopelvic CT examinations (one with STD and one with ULD protocol) for suspected focal liver lesions were included. There were 103 men and 74 women with a mean age of 64.6 ± 14.4 (SD) (range: 19—93 years). Raw data images were reconstructed with iterative reconstruction. Dose length product (DLP) and effective dose for both protocols were compared. Images were independently evaluated by two radiologists for image-quality, diagnostic quality, and confidence level. Results: DLP for STD and ULD were respectively 215.4 ± 92.0 (SD) mGy·cm (range: 76—599 mGy·cm) and 90.7 ± 37.2 (SD) mGy·cm (range: 32—254 mGy·cm). Effective dose for STD and ULD CT were 3.2 ± 1.4 (SD) mSv (range: 1.1—9.0 mSv) and 1.4 ± 0.6 (SD) mSv (range: 0.5 to 3.8 mSv). A significant 58% dose reduction was found between the two protocols (P < 0.05).

Abbreviations: BMI, Body Mass Index; CNR, contrast-to-noise ratio; CI, confidence interval; CT, computed tomography; CTDIvol, Volume CT Dose Index; DLP, dose length product; DRL, diagnostic reference level; E, effective dose; HU, hounsfield unit; IR, iterative reconstruction; ␬, Kappa coefficient; mAseff , effective mAs; MRI, magnetic resonance imaging; R, reader; ROI, region of interest; SAFIRE, sinogram affirmed iterative reconstruction; Se, sensitivity; SNR, signal-to-noise ratio; Sp, specificity; STD, standard dose; ULD, ultra low dose. ∗ Corresponding author. E-mail address: [email protected] (A. Larbi). https://doi.org/10.1016/j.diii.2017.11.003 2211-5684/© 2017 Editions franc ¸aises de radiologie. Published by Elsevier Masson SAS. All rights reserved.

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

2

A. Larbi et al. Noise, signal-to-noise ratio and contrast-to-noise ratio were higher with the ULD protocol compared to the STD protocol. No differences in subjective image quality were found between the two protocols. STD CT revealed focal liver lesions in 80 patients and ULD CT in 70 patients (P < 0.05). ULD protocol resulted in a sensitivity of 83.8% and a specificity of 96.9% for the diagnosis of focal liver lesions although it was not able to characterize them properly (Se 62.5%). Conclusion: STD CT helps detect and characterize focal liver lesions. ULD CT offers good performance to detect focal liver lesions but with lower performances for lesion characterization. © 2017 Editions franc ¸aises de radiologie. Published by Elsevier Masson SAS. All rights reserved.

Introduction During recent years, the use of computed tomography (CT) has been subjected to a dramatic increase [1], with an estimated 85 million CT studies performed in the USA in 2011. About one third of CT examinations are abdominopelvic CT ones [2]. This extensive use of CT has raised concerns about radiation-induced cancer [3]. CT contributes to half of all radiation exposure in medical practices and about a quarter of the radiation exposure per capita in the USA. Moreover, CT-based lifetime cancer risk has been estimated to be higher than 2% [1]. Due to the tremendous impact of CT on diagnosis and management, the delivered dose is now an actual matter of concern. Many methods have been employed to reduce dose delivered to patients, but these procedures may increase noise and compromise the image quality [4—8]. Hence, numerous noise-reducing techniques have been developed, such as the iterative reconstruction (IR) methods [9,10]. In recent years, IR algorithms, including sinogram affirmed ® iterative reconstruction (SAFIRE , Siemens Healthineers, Forchheim, Germany) have been developed [11]. Such IR methods provide substantial dose reduction, without compromising image quality, compared to standard dose (STD) imaging protocols [12—18]. Given that IR techniques are now widely used, it is of importance to quantify how low dose can be further reduced without compromising image quality or the diagnostic confidence. This issue is particularly relevant for neoplastic patients who undergo regular CT examinations and whose life expectancy has improved, potentially putting them at risk of radiation-induced cancer. For these patients, radiology reports have a great impact on decision-making. The purpose of this study was to evaluate and compare the diagnostic accuracy of ULD CT with that of STD CT in the detection and characterization of focal liver lesions in neoplastic patients.

Materials and methods Patients This study was approved by our institutional review board and written informed consents were waived. Participants included neoplastic patients with suspected focal liver lesions. Patients without standard of reference, those with

a body mass index (BMI) > 30 and those younger than 18 year were excluded. Patients with contraindication to intravenous administration of iodinated contrast material were excluded. The flow chart of the study design is shown in Fig. 1. Between November 2014 and January 2015, we enrolled a consecutive cohort of 177 patients with a mean age of 64.6 ± 14.4 (standard deviation [SD]) years (range: 19—93 years) and a mean BMI of 25.2 ± 8.6 (SD) kg/m2 (range: 15.0—36.9 kg/m2 ). There were 74 women (mean age = 64.9 ± 12.3 [SD] years, range: 37—91 years; mean BMI = 24.5 ± 4.3 [SD] kg/m2 , range: 15.0—35.4 kg/m2 ) and 103 men (mean age = 64.3 ± 15.8 [SD] years, range: 19—93 years; mean BMI = 25.8 ± 10.7 [SD] kg/m2 , range: 17.9—36.9 kg/m2 ).

CT protocols ®

Images were acquired with a Somatom Definition AS+ (Siemens Healthineers, Forchheim, Germany) CT unit. The detector acquisition mode was 128 × 0.6 mm2 , which corresponds to a physical collimation of 64 × 0.6 mm2 and use of a z-flying focal spot technique that allowed for double sampling along the z-direction. CT acquisitions were performed with the following parameters: data collection diameter 500-mm, 0.8 helical pitch, 0.5 s rotation time and 1-mm acquisition thickness. The tube potential was fixed to 100 kV ® and the automatic tube voltage selection (Care kV ) was activated on the cursor ‘‘parenchyma injected’’. The auto® matic tube current modulation (CareDose 4D) was used with a reference tube current of 125 mAs for the STD protocol and 55 mAs for the ULD protocol. For both protocols, raw data were reconstructed using ® the level 3 of SAFIRE . Images were reconstructed to 3 mm of thickness in transverse plane. The reconstruction kernel ‘‘moderately smooth’’ (I30f) was used. All patients underwent two abdominopelvic helical acquisitions: one with the STD protocol and the other with the ULD protocol. For 89 patients, the STD helical acquisition was performed before the ULD protocol, whereas for 88 patients, the protocols were reversed. Scanning for each acquisition was performed during relaxed inspiration, using a craniocaudal direction and starting just above the diaphragm until the symphysis pubis. Portal phase images were obtained 70 sec after the beginning of iodinated contrast material admin® istration (iohexol, 350 mg of iodine/mL, Omnipaque 350 , General Electric Healthcare, Little Chalfront, UK), which

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

Detection and characterization of focal liver lesions

Figure 1.

3

Study design. STD indicates standard dose computed tomography. ULD CT: ultra-low dose computed tomography.

was performed at an injection rate of 3-5 mL/s, using a standard power injector.

Image analysis Quantitative analysis The objective image-quality analysis was performed by a medical physicist (J.G.) using a manufacturer workstation ® (SyngoVia , Siemens Healthineers) independently of the image review results. All images were displayed with a soft tissue window (window width 370; window level 60). Mean attenuation (average of pixels) and image noise (SD of pixels) of the liver (left and right livers), portal vein, paraspinal muscle, bladder, subcutaneous or intra-abdominal fat were measured with a circular region of interest (ROI) (1 cm2 , except for the portal vein: 0.5 cm2 ) in three slices (Fig. 2). For all measurements, size, shape and position of the ROIs were kept constant between the two protocols for each patient by applying the copy-and-paste function at the workstation. The signal-to-noise (SNR) and contrast-to-noise ratio (CNR) were calculated using the formulas (1) and (2), respectively. For the CNR, the mean attenuation (HU) and

the image noise (␴) of the bladder were used as referenced in the formula [2] SNR = CNR =

|HUROI | ROI

(1)

|HUROI − HUBladder |



2 + 2 (ROI Bladder )

(2)

2

Qualitative analysis CT images were interpreted by two independent radiologists with different experiences in abdominal imaging; a junior radiologist (C.O., R1) with three years of experience and a senior radiologist (A.L., R2) with 10 years of experience. Readers were blinded to the dose and to the other reader findings. They were instructed to examine only the liver using 3 mm transverse slice with a liver window setting (window width 150; window level 70), to circumscribe all focal liver lesions and to give a diagnosis regarding the nature of the focal liver lesion. Criteria used for characterization were lesion shape, volume, attenuation, homogeneity and perilesional anomalies. The number of lesions, the dimensions of

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

4

A. Larbi et al.

Figure 2. Quantitative analysis. Computed tomography images of the abdomen and pelvis in the transverse plane obtained at three different levels. All images are displayed with a soft tissue window (window width 350; window level 60). Image noise (standard deviation of pixels) and attenuation values (average of pixels) of the liver, paraspinal muscle and subcutaneous or intra-abdominal fat (a); portal vein (b); bladder (c) are measured with a circular region of interest (ROI) of 1 cm2 except for the portal vein (0.5 cm2 ).

the largest and smallest lesions and the lesion attenuation (mean and standard deviation) were reported. Focal liver lesions were classified as malignant (primary or secondary), benign (cyst, hemangioma, adenoma, calcification, steatosis or focal nodular hyperplasia) or indeterminate. In case of discordance between the two readers, a consensus reading was performed. For each CT image, both radiologists also assessed subjectively overall image quality (1 = not evaluable; 2 = interpretable in spite of moderate artifact or noise; 3 = fully interpretable with mild noise or artifact; 4 = no artifact or noise), diagnostic quality (1 = unacceptable; 2 = suboptimal; 3 = acceptable; 4 = above average; 5 = excellent) and confidence level (1 = very poor; 2 = poor; 3 = average; 4 = high; 5 = excellent) using scales from the literature [14,15]. Interpretation was performed independently at different times on manufacturer workstations ® (Syngovia , Siemens Healthineers).

Standard of reference Proof of malignancy was determined on the basis of magnetic resonance imaging (MRI) findings, the results of histopathological analysis and by documenting lesion progression or response at serial MRI or CT examinations with a 6-month or longer time interval between examinations. Progression was demonstrated by an interval increase in size at subsequent examinations or at the index examination by comparison with previous examinations. Response was defined as a decrease in size at subsequent examinations or at the index examination by comparison with previous examinations. Imaging criteria for benign focal liver lesions were based on lesion stability of on imaging on a separate CT or MRI examination performed at least six months from the date of the index examination (either before or after). Indeterminate lesion was defined as lesion, which could not be characterized by both radiologists and needed further investigations.

Dosimetry evaluation Radiation dose differences between both CT protocols were evaluated by volume of CT dose index (CTDIvol) and the dose length product (DLP). Effective dose was calculated for each CT examination by multiplying the DLP by abdominopelvic-specific conversion coefficient (0.015 mSv·mGy−1 ·cm−1 ) [19].

Statistical analysis Statistical analysis was performed using ‘Biostatgv’ (http://marne.u707.jussieu.fr/biostatgv/). The comparisons of the mAseff , CTDIvol, DLP, Effective dose, mean attenuation, noise, signal-to-noise ratio (SNR) and contrastto-noise ratio (CNR) values between the two protocols were performed using the paired Student t-test. The agreement between radiologists for the same protocol was computed with Cohen kappa test and classified as poor ( = 0.00—0.20), fair ( = 0.21—0.40), moderate ( = 0.41—0.60), good ( = 0.61—0.80), or excellent ( = 0.81—1.00). For each radiologist, the interprotocol comparisons of the scores were evaluated using the Friedman test with a significance level set at P < 0.01. In order to determine the specificity and sensitivity of ULD CT (with the lesion as statistical unit), we considered the standard dose protocol as the standard of reference. After a consensus reading (for discordant readings), sensitivity and specificity were assessed for each patient for lesion detection, lesion characterization using ‘‘benign versus malignant’’ and lesion characterization using the nature of focal liver lesions. McNemar test was used to compare the sensitivities of the two protocols. Significance was set at P < 0.05.

Results Dosimetry Mean CTDIvol was 4.6 ± 1.8 (SD) mGy (range: 1.8—11.7 mGy) for the STD protocol and 2.0 ± 0.7 (SD) mGy (range:

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

Detection and characterization of focal liver lesions 0.8—5.1 mGy) for the ULD protocol. DLP for STD and ULD were respectively 215.4 ± 92.0 (SD) mGy·cm (range: 76—599 mGy·cm) and 90.7 ± 37.2 (SD) (range: 32—254 mGy·cm). Effective doses for STD and ULD protocols were 3.2 ± 1.4 (SD) (range: 1.1—9.0 mSv) and 1.4 ± 0.6 (SD) mSv (range: 0.5—3.8 mSv), respectively. The 58% dose reduction between the two protocols was significant (P < 0.05).

Quantitative analysis/objective image quality Quantitative results are shown in Table 1. There were no significant differences in mean attenuation between STD and ULD protocols, except for the portal vein. Noise was significantly higher (P < 0.05) with the ULD (the increase varying from 44% for the portal vein to 58% for the bladder). CNR and SNR in ULD images were significantly lower than those obtained with STD protocol (P < 0.05), regardless of the ROI (31 to 42% reduction).

Qualitative image analysis The subjective image quality for the 177 patients is shown in Table 5. With regard to STD protocol, both radiologists

Table 1

5 considered the image quality ‘‘without artifact or noise’’ or ‘‘fully interpretable’’ in the majority of patients (99% (175/177) for R2 and 95% (168/177) for R1). For ULD images, R2 scored ‘‘without artifact or noise’’ for 54% (96/177) of patients, ‘‘fully interpretable’’ for 39% (69/177) and ‘‘interpretable in spite of moderate artifact or noise’’ for 6% (12/177) whereas R1 ranked these 36% (65/177), 46% (82/177) and 16% (28/177), respectively. The interobserver agreements were excellent for STD ( = 0.91) and good for ULD ( = 0.71). Both radiologists found that diagnostic image quality was ‘‘excellent’’ or ‘‘above average’’ for STD images (94% [166/177] for R2 and 88% [156/177] for R1). For ULD images, R2 scored ‘‘excellent’’ for 24% (42/177), ‘‘above average’’ for 52% (92/177); and ‘‘acceptable’’ for 23% (41/177). R1 found 33% (59/177), 37% (66/177) and 21% (38/177) respectively. The interobserver agreements were excellent for STD ( = 0.89) and good for ULD ( = 0.76). The diagnostic confidence level was ‘‘excellent’’ or ‘‘high’’ for all patients for both radiologists for STD images (98% [173/177] for R2 and 90% [160/177] for R1). For ULD images, R2 scored ‘‘excellent’’ for 59% (105/177), ‘‘high’’ for 34% (60/177) and ‘‘average’’ for 6% (11/177). R1 found 38% (67/177), 37% (65/177) and 18% (32/177). The interobserver agreements were good for STD ( = 0.71) and ULD ( = 0.66).

Quantitative analysis and comparison between two CT protocols.

Mean attenuation (HU) Left liver Right liver Muscle Fat Portal vein Bladder Noise (HU) Left liver Right liver Muscle Fat Portal vein Bladder SNR Left liver Right liver Muscle Fat Portal vein Bladder CNR Left liver Right liver Muscle Fat Portal vein

STD

ULD

% diff

P value

110.2 ± 21 101.8 ± 21.2 50.4 ± 13.9 −98.7 ± 15.7 194.1 ± 35.3 11.5 ± 9.7

108.7 ± 20 102.2 ± 18.8 50.5 ± 14.9 −98.6 ± 15.8 171.2 ± 30 11.7 ± 10.3

−1 0 0 0 −12 2

0.06 0.05 0.77 0.87 < 0.05 0.60

16.4 ± 2.9 16.9 ± 3.2 16.1 ± 3.2 17.3 ± 3.6 19.6 ± 4 14.5 ± 3

24.1 ± 4 25.6 ± 4.8 23.9 ± 4.4 25 ± 4.7 28.3 ± 5.7 22.9 ± 5.2

47 51 48 45 44 58

< 0.05 < 0.05 < 0.05 < 0.05 < 0.05 < 0.05

6.7 ± 2 6.0 ± 2 3.1 ± 1.3 5.7 ± 1.7 9.9 ± 2.3 0.8 ± 0.7

4.6 ± 1.3 4.0 ± 1.1 2.1 ± 0.8 3.9 ± 1.1 6.9 ± 1.5 0.5 ± 0.4

−32 −34 −33 −31 −31 −37

< 0.05 < 0.05 < 0.05 < 0.05 < 0.05 < 0.05

6.4 ± 2 5.7 ± 2 2.5 ± 1.1 6.9 ± 1.6 10.6 ± 2.7

4.1 ± 1.6 3.7 ± 1.5 1.7 ± 1 4.6 ± 1.4 6.2 ± 2

−35 −35 −35 −33 −42

< 0.05 < 0.05 < 0.05 < 0.05 < 0.05

Values are expressed in mean ± standard deviation. STD: standard dose; ULD: ultra-low dose; % diff: difference between the image quality indexes for STD and ULD protocols.

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model

ARTICLE IN PRESS

DIII-1004; No. of Pages 10

6

A. Larbi et al. Table 2

Total number of focal liver lesions detected for each patient with two CTprotocols. STD

Number of lesions

STD vs ULD

Number of patients R1

0 1-5 6-10 11-20 20-30

ULD

106 60 2 5 4

R2 97 70 2 3 5

R1 107 59 3 5 3

R2 107 59 3 5 3

STD vs ULD

P-value

Interprotocol

Kappa

Interobserver

R1

R2

R1

R2

0.083 0.083 0.083 0.083 0.083

P < 0.05 P < 0.05 P < 0.05 P < 0.05 P < 0.05

0.88 0.88 0.88 0.88 0.88

1

R1 and R2 correspond to the reader 1 (CO) and 2 (AL), respectively. STD: standard dose; ULD: ultra-low dose.

For all subjective image-quality criteria, the interprotocol comparisons showed significant differences for both radiologists (P < 0.05).

(95% CI: 91—99%), respectively (P = 6.0 × 10−6 ) (Table 4) (Figs. 3, 4).

Lesion detection

Discussion

The total number of focal liver lesions detected for each patient with the two protocols is presented in the Table 2. After consensus reading, a total of 346 focal liver lesions in 85 patients (92 patients did not have focal liver lesions) were considered present using the standard of reference. Using STD protocol CT, R1 detected 336 focal liver lesions in 71 patients (71/85; 83.5%) and R2 351 focal liver lesions in 80 patients (80/85; 94.1%). Using ULD CT, R1 and R2 detected 269 and 278 focal liver lesions in 70 patients (70/85; 82.4%), respectively (Table 2). Concerning the ULD protocol, sensitivity for focal liver lesions detection (with STD as gold standard) was 83.8% (95% CI: 74—90%). The 16.2% difference in sensitivity on a perpatient basis between ULD and STD protocol was significant (P = 0.024). Specificity was 96.9% (95% CI: 91—99%) (Table 4).

The present study suggests that there is no clinically relevant difference in lesion detection between ULD and STD CT protocols. ULD CT, with a dose lower than abdominal X-ray, can be performed to detect focal liver lesions in neoplastic patients, but not for lesion characterization. Since 2012, IR protocols have been fully incorporated into routine clinical practice in our institution [10,14]. Indeed, in recent years, IR has been shown to allow substantial dose reduction without compromising image quality thus allowing the use of reduced-dose CT imaging protocols [13,16—18,20,21]. With an optimized STD protocol, we have yet to determine the lowest level of dose reduction that can be used with acceptable performances in terms of detection and diagnosis. So, we chose to first evaluate focal liver lesions in neoplastic patients as we have access to a relatively large number of patients in our institution and thus a sufficient number of lesions. As these patients undergo abdominal CT sometimes as often as four times a year and have a longer life expectancy, reducing radiation dose is a matter of concern in this population. As no clear standard has been defined concerning ULD abdominal CT, we aimed at obtaining CT dose lower than those of abdominal X-ray, as it has previously been done for chest CT and chest X-ray [13]. ® The STD protocol, using SAFIRE , allows a substantial dose reduction. Dose levels (CTDIvol = 4.6 ± 1.8 (SD) mGy (range: 1.8—11.7 mGy) are equivalent to low dose protocols and lower than STD protocols in studies that compared IR protocols with filtered-back projection ones [20—26]. The delivered dose for this protocol was significantly lower than the National DRL (73% reduction). Dose reduction usually lessens image quality; thus, IR was used to balance this reduced quality and maintain acceptable diagnostic confidence level [14]. Indeed, IR may achieve lower doses than conventional radiology studies while maintaining quality of image suitable for diagnostic purposes [27]. Our ULD protocol result in an effective dose of 1.4 ± 0.6 mSv, that is lower than the National DRL for abdominal X-ray (effective dose: 1.82 mSv; dose area product: 7000 mGy·cm2 ; and conversion factor:

Lesion characterization Using the standard reference, 85 patients (85/177; 48%) had 346 focal liver lesions, benign or malignant. One patient (1/85; 1.2%) had primary liver cancer, 25 patients (25/85; 29.4%) had liver metastasis, 59 (59/85; 69.4%) had benign lesions and 92 (92/177; 52%) had no lesions. Of the 26 patients with malignant liver lesions, 10 (10/26; 38.4%) had histological confirmation after liver biopsy. The nature of the focal liver lesions is shown in Table 3. Benign lesions were classified as calcifications (n = 7), hemangioma (n = 1), cysts (n = 42) and focal steatosis (n = 4). Characterization was not achievable for 11 focal liver lesions by either protocol and classified as indeterminate. After further investigations, these were cysts (n = 2), hemangioma (n = 1), focal steatosis (n = 2) and metastases (n = 5). When considering focal liver lesion characterization using ‘‘benign versus malignant’’, sensitivity and specificity of ULD (with STD as reference standard protocol) were 66.3% (95% CI: 55—76%) and 96.9% (95% CI: 91—99%), respectively (P = 2.7 × 10−5 ). When distinguishing between the types of focal liver lesions, sensitivity and specificity for characterization were 62.5% (95% CI: 52—72%) and 96.9%

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model

ARTICLE IN PRESS

DIII-1004; No. of Pages 10

Detection and characterization of focal liver lesions Table 3

7

Focal liver lesions characterization with two CT protocols. STD

Malignant lesions Primary Metastatic Benign lesions Calcification Hemangioma Cyst(s) Steatosis Indeterminate Present

ULD

STD vs. ULD

R1 vs. R2

P-value

Interprotocol

Kappa

Interobserver

R1

R2

R1

R2

R1

R2

STD

ULD

0 27

1 25

1 19

1 19

0.083 0.083

0.083 0.083

0.96 0.96

1 1

7 1 42 4

7 1 42 4

5 1 29 2

6 1 31 2

P < 0.05 P < 0.05 P < 0.05 P < 0.05

P < 0.05 P < 0.05 P < 0.05 P < 0.05

1 1 1 1

0.94 0.94 0.94 0.94

11

11

6

7

0.259

0.416

1

0.92

R1 and R2 correspond to reader 1 (CO) and 2 (AL), respectively. STD: standard dose; ULD: ultra-low dose. Indeterminate focal liver lesion was defined as lesion, which could not be characterized by both radiologists and needed further investigations.

Table 4 Sensitivity and specificity for detection and characterization of focal liver lesions for ULD protocol on a perpatient basis. Sensitivity (%)

Lesion detection FN FP VN VP Benign versus malignant FN FP VN VP Nature of lesionsa FN FP VN VP

Specificity (%)

Value

Estimate

95% CI

P-value

Estimate

95% CI

13 3 94 67

84

[74-90]

0.024

97

[91-99]

27 3 94 53

66

[55- 76]

2.7 × 10−5

97

[91-99]

30 3 94 50

63

[52-72]

6.0 × 10−6

97

[91-99]

CI: confidence interval; FN: false negative; FP: false positive; TN: true negative; TP: true positive. a Calcifications, hemangioma, cysts and focal steatosis.

2.6 × 10−4 mSv/mGy·cm2 ) [28]. There was a 58% dose reduction between our two protocols and 88% dose reduction between our ULD and the National NRD. Concerning objective image quality, there were higher levels of noise, which is not surprising given the lower dose levels and no increase in the level of SAFIRE [22,25,26]. Shuman et al. found a similar SNR but lower CNR [20]. Both readers mostly classified subjective diagnosis image quality as above average or excellent, with a high or excellent confidence level; suggesting higher levels of noise were acceptable for diagnosis purpose. There were no differences in subjective image quality parameters, despite higher image noise and lower CNR and SNR.

Compared to the STD protocol, ULD protocol resulted in a sensitivity of 83.8% for the detection of focal liver lesions in this study. This result suggests, consistently with previous studies, that objective image quality parameters do not necessarily correlate with diagnostic performance. Kuszyk et al. found a sensitivity of 81% of portal phase CT for the detection of focal liver lesions [29]. Vialle et al. found a sensitivity of 80% for detection of hepatic metastases [30]. For lesion characterization, the ULD protocol was inferior to STD CT studies, with a sensitivity as low as 62.5% compared to 77—87% in the literature [26,31]. This result can be explained by increased noise resulting in a reduction of contrast between the liver parenchyma and the liver lesion.

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

8

ARTICLE IN PRESS A. Larbi et al.

Figure 3. Contrast-enhanced CT image of the liver in a 59-year-old woman with liver metastasis from lung cancer. The two readers correctly detected and characterized liver metastases with the STD (a—c) and ULD (b—d) protocol with a soft tissue- (a—b) (window width 370; window level 60) (white arrows) and a liver (c—d) (window width 150; window level 70) (grey arrows) windows setting.

Figure 4. Contrast-enhanced CT image of the liver in a 61-year-old woman with focal liver lesion. The two readers correctly detected a liver nodule (arrows) with the STD (a—c) and ULD (b—d) protocol whereas they were able to characterize the lesion (a—c), which was a simple cyst only with the STD protocol.

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

Detection and characterization of focal liver lesions Our results have several clinical implications. They emphasize the fact that, while dose reduction is a priority, it must not be obtained at the penalty of lower degrees of sensitivity. ULD CT can be an option in terms of liver lesion detection, but it is limited for lesion characterization and needs further MRI examination for correct lesion categorization. Further studies are needed to evaluate intermediate levels of dose reduction and determine the optimal level for patient care. This study had some limitations. First, we only analyzed portal phase abdominal CT examinations and not multiphasic ones. We consider that in clinical routine, especially in neoplastic patients, radiologists often perform only portal phase CT, with the exception of hepatocellular carcinoma or other hypervascular lesions [32]. However, the vast majority of lesions in our population were metastatic or benign ones and patients were not referred for the characterization of liver lesions. Second, our study has only a limited number of patients with a strong standard of reference, such as pathologic examination, which also reflects daily clinical routine, where proof of malignancy in metastatic disease is rarely definitely confirmed by histopathological analysis. The standard reference was established on the basis of MRI examinations and a review of clinical data and multimodality imaging patient follow-up at six months. Such reference represents the best standard of reference available in the absence of systematic histologic proof. Third, we chose not ® to increase the level of SAFIRE , which would have resulted in an over-smooth appearance; or to increase image thickness. Finally, we only compared one particularly low level of dose reduction to our standard of care. Further studies would compare less dramatic dose reduction to determine the optimum dose. In conclusion, our results suggest that neoplastic patients can undergo serial STD CT examinations for the initial staging of the disease and then the follow-up could be done using an ULD CT examination for detecting liver lesion. If liver lesions are detected, further MRI examinations should be necessary to characterize these lesions.

Acknowledgement We are grateful to Sarah Kabani for editing the manuscript.

Disclosure of interest The authors declare that they have no competing interest.

References [1] Brenner DJ, Hall EJ. Computed tomography: an increasing source of radiation exposure. N Engl J Med 2007;357:2277—84. [2] Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med 2009;169:2078—86. [3] Berrington de Gonzalez A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med 2009;169:2071—7.

9 [4] Kalra MK, Maher MM, Toth TL, et al. Strategies for CT radiation dose optimization. Radiology 2004;230:619—28. [5] Gies M, Kalender WA, Wolf H, Suess C. Dose reduction in CT by anatomically adapted tube current modulation. I. Simulation studies. Med Phys 1999;26:2235—47. [6] Brooks RA, Di Chiro G. Theory of image reconstruction in computed tomography. Radiology 1975;117:561—72. [7] Kalender WA, Wolf H, Suess C. Dose reduction in CT by anatomically adapted tube current modulation. II. Phantom measurements. Med Phys 1999;26:2248—53. [8] Kalra MK, Maher MM, Blake MA, et al. Detection and characterization of lesions on low-radiation-dose abdominal CT images postprocessed with noise reduction filters. Radiology 2004;232:791—7. [9] Patino M, Fuentes JM, Singh S, Hahn PF, Sahani DV. Iterative reconstruction techniques in abdominopelvic CT: technical concepts and clinical implementation. AJR Am J Roentgenol 2015;205:W19—31. [10] Greffier J, Fernandez A, Macri F, Freitag C, Metge L, Beregi JP. Which dose for what image? Iterative reconstruction for CT scan. Diag Interv Imaging 2013;94:1117—21. [11] Siemens AGS. Somatom definition as user manual. Germany; 2012. [12] Macri F, Greffier J, Pereira FR, Mandoul C, Khasanova E, Gualdi G, et al. Ultra-low-dose chest CT with iterative reconstruction does not alter anatomical image quality. Diagn Interv Imaging 2016;97:1131—40. [13] Macri F, Greffier J, Pereira F, et al. Value of ultra-low-dose chest CT with iterative reconstruction for selected emergency room patients with acute dyspnea. Eur J Radiol 2016;85:1637—44. [14] Greffier J, Macri F, Larbi A, Fernandez A, Pereira F, Mekkaoui C, et al. Dose reduction with iterative reconstruction in multi-detector CT: what is the impact on deformation of circular structures in phantom study? Diagn Interv Imaging 2016;97:187—96. [15] Gervaise A, Gervaise-Henry C, Pernin M, Naulet P, JuncaLaplace C, Lapierre-Combes M. How to perform low-dose computed tomography for renal colic in clinical practice. Diagn Interv Imaging 2016;97:393—400. [16] Mitsumori LM, Shuman WP, Busey JM, Kolokythas O, Koprowicz KM. Adaptive statistical iterative reconstruction versus filtered back projection in the same patient: 64 channel liver CT image quality and patient radiation dose. Eur Radiol 2012;22:138—43. [17] Blum A, Gervaise A, Teixeira P. Iterative reconstruction: why, how and when? Diagn Interv Imaging 2015;96:421—2. [18] Burckel LA, Defez D, Chaillot PF, Douek P, Boussel L. Use of an automatic recording system for CT doses: evaluation of the impact of iterative reconstruction on radiation exposure in clinical practice. Diagn Interv Imaging 2015;96:265—72. [19] Shrimpton PC, Wall BF, Yoshizumi TT, Hurwitz LM, Goodman PC. Effective dose and dose—length product in CT. Radiology 2009;250:604—5. [20] Shuman WP, Chan KT, Busey JM, et al. Standard and reduced radiation dose liver CT images: adaptive statistical iterative reconstruction versus model-based iterative reconstruction-comparison of findings and image quality. Radiology 2014;273:793—800. [21] Pickhardt PJ, Lubner MG, Kim DH, et al., Abdominal CT. with model-based iterative reconstruction (MBIR): initial results of a prospective trial comparing ultralow-dose with standard-dose imaging. AJR Am J Roentgenol 2012;199:1266—74. [22] Bodelle B, Isler S, Scholtz JE, et al. Benefits of sinogramaffirmed iterative reconstruction in 0.4 mSv ultra-low-dose CT of the upper abdomen following transarterial chemoembolisation: comparison to low-dose and standard-dose CT and filtered back projection technique. Clin Radiol 2016;71:e11—5. [23] Matsuki M, Murakami T, Juri H, Yoshikawa S, Narumi Y. Impact of adaptive iterative dose reduction (AIDR) 3D on low-dose

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003

+Model DIII-1004; No. of Pages 10

ARTICLE IN PRESS

10

[24]

[25]

[26]

[27]

A. Larbi et al. abdominal CT: comparison with routine-dose CT using filtered back projection. Acta Radiol 2013;54:869—75. Volders D, Bols A, Haspeslagh M, Coenegrachts K. Model-based iterative reconstruction and adaptive statistical iterative reconstruction techniques in abdominal CT: comparison of image quality in the detection of colorectal liver metastases. Radiology 2013;269:469—74. Deak Z, Grimm JM, Treitl M, et al. Filtered back projection, adaptive statistical iterative reconstruction, and a model-based iterative reconstruction in abdominal CT: an experimental clinical study. Radiology 2013;266: 197—206. Kamel IR, Choti MA, Horton KM, et al. Surgically staged focal liver lesions: accuracy and reproducibility of dualphase helical CT for detection and characterization. Radiology 2003;227:752—7. McLaughlin PD, Ouellette HA, Louis LJ, et al. The emergence of ultra-low — dose computed tomography and the impending obsolescence of the plain radiograph? Can Assoc Radiol J 2013;64:314—8.

[28] European Commission, European guidance on estimating population doses from medical x-ray procedures radiation protection. European Commission, DG Energy-Transport; 2008. p. 154. [29] Kuszyk BS, Bluemke DA, Urban BA, et al. Portal-phase contrastenhanced helical CT for the detection of malignant hepatic tumors: sensitivity based on comparison with intraoperative and pathologic findings. AJR Am J Roentgenol 1996;166:91—5. [30] Vialle R, Boucebci S, Richer JP, et al. Preoperative detection of hepatic metastases from colorectal cancer: prospective comparison of contrast-enhanced ultrasound and multidetector-row computed tomography (MDCT). Diagn Interv Imaging 2016;97:851—5. [31] van Kessel CS, van Leeuwen MS, van den Bosch MA, et al. Accuracy of multislice liver CT and MRI for preoperative assessment of colorectal liver metastases after neoadjuvant chemotherapy. Dig Surg 2011;28:36—43. [32] Cassinotto C, Aubé C, Dohan A. Diagnosis of hepatocellular carcinoma: an update on international guidelines. Diagn Interv Imaging 2017;98:379—91.

Please cite this article in press as: Larbi A, et al. Detection and characterization of focal liver lesions with ultra-low dose computed tomography in neoplastic patients. Diagnostic and Interventional Imaging (2018), https://doi.org/10.1016/j.diii.2017.11.003