Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection

Physica Medica xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Physica Medica journal homepage: http://www.physicamedica.com

Original paper

Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection Jijo Paul a,⇑, Annamma Chacko b, Paola Saccomandi c, Thomas J. Vogl a, Nour-Eldin A. Nour-Eldin a,d a

Department of Diagnostic and Interventional Radiology, University Hospital, Johann Wolfgang Goethe University, Theodor-Stern-Kai 7, 60590 Frankfurt, Germany Division of Applied Mathematics and Statistics, Dougherty System, 31705 Albany, GA, United States c Department of Biomedical Instrumentation, Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21, 00128 Rome, Italy d Department of Diagnostic and Interventional Radiology, University Hospital, Cairo University, Cairo, Egypt b

a r t i c l e

i n f o

Article history: Received 21 September 2016 Received in Revised form 14 November 2016 Accepted 20 November 2016 Available online xxxx Keywords: Cone-beam computed tomography Image acquisition protocols Patient dose Lesion delineation Lesion detection Image quality

a b s t r a c t Objective: To evaluate the effect of cone-beam computed tomography (CBCT) image acquisition protocols on image quality, lesion detection, delineation, and patient dose. Methods: 100-patients and a CTDI phantom combined with an electron density phantom were examined using four different CBCT-image acquisition protocols during image-guided transarterial chemoembolization (TACE). Protocol-1 (time: 6 s, tube rotation: 360°), protocol-2 (5 s, 300°), protocol-3 (4 s, 240°) and protocol-4 (3 s, 180°) were used. The protocols were first investigated using a phantom. The protocols that were found to be clinically appropriate in terms of image quality and radiation dose were then assessed on patients. A higher radiation dose and/or a poor image quality were inappropriate for the patient imaging. Patient dose (patient-entrance dose and dose-area product), image quality (Hounsfield Unit, noise, signal-to-noise ratio and contrast-to-noise ratio), and lesion delineation (tumor-liver contrast) were assessed and compared using appropriate statistical tests. Lesion detectability, sensitivity, and predictive values were estimated for CBCT-image data using pre-treatment patient magnetic resonance imaging. Results: The estimated patient dose showed no statistical significance (p > 0.05) between protocols-2 and -3; the assessed image quality between these protocols manifested insignificant difference (p > 0.05). Two other phantom protocols were not considered for patient imaging due to significantly higher dose (protocols-1) and poor image quality (protocol-4). Lesion delineation and detection were insignificant (p > 0.05) between protocols-2 and -3. Lesion sensitivities generated were 81–89% (protocol-2) and 81–85% (protocol-3) for different lesion types. Conclusion: Data acquisition using protocols-2 and -3 provided good image quality, lesion detection and delineation with acceptable patient dose during CBCT-imaging mainly due to similar frame numbers acquired. Ó 2016 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.

Key points
Effect of CBCT image acquisition protocols on patient doses and associate image quality are scarce in literature.
Recent CBCT device capable to produce multiple imaging protocols with different rotation angles and frame numbers.
Data acquisition protocol-2 and -3 produced good image quality, lesion detection and delineation with acceptable patient dose. ⇑ Corresponding author. E-mail addresses: [email protected] (J. Paul), sajinichacko2011@gmail. com (A. Chacko), [email protected] (P. Saccomandi), T.Vogl@ em-uni-frankfurt.de (T.J. Vogl), [email protected] (N.-E.A. Nour-Eldin).

Introduction Diagnostic cone-beam computed tomography (CBCT) imaging produces patient cross-sectional image data, which can display patient internal anatomy, vascular and lesion information with contrast material (CM) [1]. An accurate three-dimensional tissue characterization is essential to perform many interventional procedures, locate anatomical structures, blood vessels and embedded parenchymal lesions [2]. Many clinical studies have shown the usability of diagnostic CBCT during interventional procedures [2–6]. Truncated projections and beam hardening effects are common problems associated with CBCT imaging that significantly affects reconstruction of the acquired image data. Several studies have

http://dx.doi.org/10.1016/j.ejmp.2016.11.111 1120-1797/Ó 2016 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

2

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

shown that the production of scatter radiation in the imaging field directly affects image quality and patient dose [7–9]. The scatter radiation increased patient dose and reduced image quality during CBCT patient imaging. Higher radiation doses during CBCT imaging and increased frequency of examinations performed increases the need to reduce radiation dose to the patient and working staff. To maintain consistent performance of CBCT equipment, the European Federation of Organizations for Medical Physics (EFOMP) working group conducted discussions since 2014 and formulated certain guidelines [10]. These guidelines provided a minimum number of quality tests which needed to be performed in order to ensure safety, reliability, and consistency in the operation of the device. Recent CBCT imaging devices are capable of being used for multiple image acquisition protocols with different rotation angles and frame numbers for patient examinations during image-guided interventional procedures. To our knowledge there are few publications available that discuss several clinical CBCT image acquisition protocols. Image quality improvement, and lesion detection using multiple acquisition protocol increases the concern regarding patient dose. Present study evaluated several CBCT image acquisition protocols and its impacts on lesion detection, delineation, image quality and patient dose during image-guided hepatic transarterial chemoembolization (TACE) therapy.

Material and methods Study design Patients and phantom imaging were performed in our University hospital using a robotic multi-axis Artis zeego CBCT imaging system from Siemens Healthcare, Forchheim, Germany. Institutional review board approval was obtained prior to beginning the study. The CBCT protocols were first investigated using a phantom. Next, the protocols that were found to be clinically appropriate in terms of image quality and radiation dose were then assessed on patients. Phantom imaging and protocols Phantom imaging was used to evaluate different image acquisition protocols, which are available with the CBCT imaging system to demonstrate patient radiation dose optimization and safety. A 32 cm diameter computed tomographic dose index (CTDI) body phantom (14 cm length; PTW, Freiburg, Germany) combined with an electron density phantom (5 cm length; Gammex, model 467, Middleton, USA; Fig 1) to cover longitudinal axis of the exposure field, was used for CBCT imaging. Specifications of the electron density phantom and inserted rod materials can be obtained from the [11]. The imaging was performed using a single rotation X-ray tube-detector system around the phantom in a pre-determined software controlled angle. The phantom imaging was repeated five times for each acquisition protocol to generate and report stable results. The protocols used for phantom imaging were as follows: protocol-1 (6 s imaging time, 360° rotation angle), protocol-2 (5 s, 300°), protocol-3 (4 s, 240°) and protocol-4 (3 s, 180°). Furthermore, additional information regarding the acquisition protocols used for imaging is provided in Table 1. A modified FDK (Feldkamp-Davis-Kress) algorithm was used to reconstruct the cross-sectional image data. Reconstructed image slice thickness was 0.7 mm and the software version used for imaging was VC 21A (Siemens Healthcare, Forchheim, Germany). Patient selection

Fig. 1. (A) Pictorial representation of the phantom cross-section used for conebeam CT imaging to assess radiation dose and image quality with different image data acquisition protocols.

One hundred patients’ imaging data acquired in the past two years was retrospectively analyzed. Patient CBCT examinations were performed exclusively for image-guidance purpose during real-time TACE therapy and the selection of patients was random. Patients with hepatic lesions contraindicated for surgery or unresponsive lesions to chemotherapy were included in this study. Patients with lesion sizes less than 1 cm diameter and appearance of >5 hepatic lesions in hepatic parenchyma were excluded from the present investigation.

Table 1 Description of the cone-beam CT image acquisition protocols used for the phantom imaging. Imaging parameters

Kilo-voltage (kV) Tube current (mA) Exposure time (s) Imaging start position (°) Imaging end position (°) Rotation speed (°/s) Total rotation angle (°) Frame rate (s) Number of frames

Phantom imaging Protocol-1

Protocol-2

Protocol-3

Protocol-4

124 333 ± 3 (329–335) 6 0 360 60 360 60 397

124 340 ± 3 (331–347) 5 0 300 60 300 60 248

124 360 ± 4 (357–363) 4 0 240 60 240 60 248

124 377 ± 4 (375–379) 3 0 180 60 180 60 166

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

3

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

Patient MR-imaging Initial evaluations of hepatic lesions were performed using pretreatment magnetic resonance image (MRI) data by the procedural personnel, as this imaging modality is considered to be of gold standard when assessing hepatic lesions. Contrast enhanced T1weighted and T2-weighted MR-image data were acquired to assess lesion details (Table 2). A 1.5-Tesla Magnetom Symphony MRIsystem from Siemens Healthcare, Germany was used to perform MRI examinations. Patient cone-beam CT imaging and protocol Patients were positioned supine on the table and hands were placed on the head during CBCT imaging. Two acquisition clinical protocols (protocol-2 and -3) were used for patient imaging. Fifty patients (mean age 62 ± 15 (46–78); male–female ratio 30:20) were imaged using protocol-2, and consisted of 5 s imaging time with 300° rotation angle. The remaining fifty patients (mean age 61.5 ± 12 (47–79); male–female ratio 28:22) were imaged using protocol 3 (4 s time with 240° rotation angle). Both protocols 2 and -3 scan settings were the same as the phantom imaging except for kV (96.3 ± 5 (91–104)) and mA (411.6 ± 39 (389–448)). A preset 90 kV tube voltage was used for patient imaging; however, kV automatically modulated with tube current during exposure to maintain a constant detector-entrance dose (0.36 lGy/ image). 12 mL of Visipaque 320 (3.84 g of iodine) contrast material (GE Healthcare, Braunschweig, Germany) injection was performed at the hepatic artery using a 2.7F/2.4Fx150 cm co-axial catheter

(Trevo Pro 18 micro-catheter, Concentric Medical, CA, USA). After image data acquisition, the axial images were reconstructed using a Siemens X-Leonardo workstation. The reconstructed images were then exported to RIS/ PACS network system (Centricity 4.1, GEHealthcare, Dornstadt, Germany) for analysis. Radiation dose assessment Dose metrics used to report patient and phantom dose were the patient-entrance-dose (PED) and dose-area-product (DAP) [12]. PED and DAP are the vendor prescribed dose descriptors representing radiation dose received by patients during CBCT examination. Each acquisition protocol generated X-ray radiation dose readings during imaging and they were safely saved in the PACS network system. DAP values were generated using a properly calibrated DAP meter manufactured by PTW (model K1S), Freiburg, Germany. A DAP meter was affixed at the emergence point of the radiation beam from the X-ray tube housing. CBCT system automatically computes PED for each patient and phantom examinations using certain reference conditions published by International Electrotechnical Commission (IEC) [13]. Comparisons of phantom dose parameters among different protocols were presented as percentage deviation (%). Mean ± standard deviation (SD) and range was used to represent numerical values of the continuous variables. Statistical significance between image acquisition protocols represented as p-value and a pvalue 6 5% (0.05) was considered to be statistically significant differences. Shapiro-Wilk statistical test was used to examine the normality of data distribution. Statistical analyses were performed

Table 2 Patient tumor characteristics extracted from gold standard MRI cross-sectional image data, which shows the patients with CBCT imaging protocols (P2 and P3) separately. (HOEL: Homogenously enhanced lesion; HEEL: Heterogeneously enhanced lesion; UEL: Unenhanced lesions; P = image acquisition protocol). Lesion category

Lesion location (P2/P3)

Clinical category

P2

P3

P2

P3

P2

P3

HOEL

Right lobe: 27/22 Left lobe: 19/19 Caudate lobe: 7/8 Quadrate lobe:12/9

Cholangio carcinoma Hepatocellular carcinoma

10 7

9 10

40 25

31 27

3.6  3 3.5  2.9

4.1  3.3 4.2  3.4

HEEL

Right lobe: 23/18 Left lobe: 21/13 Caudate lobe: 5/5 Quadrate lobe: 4/7

Hepatocellular carcinoma Cholangio carcinoma Metastasis from: Renal cell carcinoma Breast carcinoma

5 5

6 4

15 22

17 14

4.4  3.9 5.1  4.3

4.9  3.8 4.4  3.9

3 2

4 2

11 5

9 3

3.1  2.9 3.8  2.9

3.6  3.2 4.2  3.9

Right lobe: 23/20 Left lobe: 22/16 Caudate lobe: 9/1 Quadrate lobe: 7/4

Metastasis from: Sigmoid carcinoma Colon carcinoma Gastric carcinoma

6 8 4

3 10 2

21 26 14

13 23 5

3.5  2.4 3.2  2.5 3.7  2.6

3.8  3.3 3.2  2.7 2.9  2.5

UEL

Number of patients

Number of lesions

Tumor dimension in cm (l  b)

Table 3 Displays patient hepatic lesion characteristics determined using CBCT cross-sectional images. P2 and P3 are two acquisition protocols used to generate CBCT image data. Type of lesion

Lesion clinical categories

HOEL

Cholangio carcinoma: Hepatocellular carcinoma:

HEEL

Hepatocellular carcinoma: Cholangio carcinoma: Metastasis from: Renal cell carcinoma: Breast carcinoma: Metastasis from: Sigmoid carcinoma: Colon carcinoma: Gastric carcinoma:

UEL

Number of patients

Number of lesions

Length  breadth (cm)

P2

P3

P2

P3

P2

P3

P2

P3

10 7

9 10

38 23

27 24

3.4  2.9 3.6  2.8

3.8  3 4  3.3

258 (226–279) 242 (221–273)

240 (220–263) 247 (228–269)

5 5

6 4

13 19

15 12

4.2  3.9 4.9  4

4.9  3.6 4.3  3.9

49 (31–59) 44 (29–54)

43 (29–55) 47 (31–57)

3 2

4 2

9 4

7 3

2.9  2.7 3.8  2.9

3.6  3.2 4  3.7

43 (30–51) 45 (32–54)

38 (27–47) 39 (29–49)

6 8 4

3 10 2

19 24 12

12 22 4

3.6  2.4 2.9  2.3 3.7  2.5

3.8  3.2 2.9  2.5 2.9  2.4

14 (818) 9 (514) 12 (716)

11 (616) 14 (918) 8 (412)

Mean hounsfield unit

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

4

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

using computer based software BiAS-9.02, Epsilon-verlag, Darmstadt, Germany. Radiation dose parameters such as PED and DAP estimated for different acquisition protocols were compared using the Student’s t-test.

clinical experience with abdominal computed tomographic imaging.

Image quality assessment

Hounsfield Unit (HU) of the hepatic lesions and parenchyma of the anatomical structures including liver, kidney, spleen and muscle were measured. The hepatic lesions were divided into three categories: (a) homogenously enhanced lesion (HOEL), heterogeneously enhanced lesion (HEEL), and unenhanced lesion (UEL) based on the contrast material enhancement (Table 3). To estimate HU, two to three centimeter diameter region of interest (ROI) was repeatedly drawn three times on each slice for organ and hepatic lesions that were to be analyzed on three adjacent slices and the mean value was calculated. The measured position and the size of the ROI were repeated for all patient images that were used to calculate mean HU. Standard deviation of all pixels in the region of interest was ascertained and presented as image noise; furthermore, we considered all measured ROIs for the determination of image noise in an anatomical structure. Signal-to-noise ratio (SNR) was calculated using HU and image noise for anatomical locations/lesions. SNR = HU specified structure/noise. Contrast-tonoise ratio (CNR) of the anatomical structures or lesion was assessed using one of the following formulas:

Acquired patient image data was quantitatively and qualitatively analyzed by three expert radiologists, who have >7-years Table 4 Qualitative evaluation scale (5-point grading) used for the assessment of hepatic lesion delineation. Grading Score

Description

Details of the grading score used for qualitative lesion delineation 1 Inability to differentiate hepatic lesions from the parenchyma, lesion diagnosis is impossible 2 Visible hepatic lesions with unclear margins, a delineation of lesions from the parenchyma is difficult 3 Moderate differentiation of lesions from the hepatic parenchyma 4 Good differentiation of lesions from the hepatic parenchyma, good lesion delineation 5 Differentiation of hepatic lesions from the parenchyma is more than necessary

Patient image data analysis

Fig. 2. The image quality determinants such as the Hounsfield Unit (A), and the signal-to-noise ratio (B) obtained from the phantom materials while it imaged with CBCT. The HU (C) and image noise (D) obtained with different patient imaging protocols for tumor types and anatomical structures derived from the axial images.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

5

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

CNR ¼ ðHUcontrastenhancedstructure  HUmuscle Þ=imagenoise CNR ¼ ðHUlesion  HUhepaticparenchyma Þ=imagenoise Image quality parameters such as HU, image noise, SNR and CNR between different protocols were tested using a Student’s ttest. The correlations between the protocols were assessed using a Pearson correlation coefficient (R2). We further evaluated lesion conspicuity using the attenuation difference between the lesion and the hepatic parenchyma [14]. Tumor-liver-contrast

TLC ¼ ROIlesion  ROIhepaticparenchyma Patient qualitative lesion delineation assessments were performed by the same radiologists during evaluation of different hepatic lesion categories. CM injection and image acquisition protocol were blinded to the evaluators during assessment; however, they were informed regarding the examinations performed for TACE. All lesion delineation evaluations were performed on a RIS/ PACS workstation and explanation of the grading scale used for evaluation is displayed in Table 4. Estimated lesion delineation qualitative scores were compared using a non-parametric Wilcoxon rank-sum test for different acquisition protocols. Moreover, during qualitative analysis the inter-reader comparisons were carried out using Cohen’s Kappa statistical test. Kappa score represent less than chance agreement (k < 0), slight agreement (k = 0.01–0.20), fair agreement (k = 0.21– 0.40), moderate agreement (k = 0.41–0.60), substantial agreement (k = 0.61–0.80) and almost perfect agreement (k = 0.81–0.99). Hepatic lesion sensitivity and predictive values were estimated using gold standard MR image data; furthermore, MR image data was used as standard of reference for statistical evaluation during hepatic lesion detection using CBCT. Sensitivity, specificity and predictive values were calculated using MedCalc easy-to-use statistical software (Version 14.8.1). Phantom image data analysis The HU values were quantitatively measured using ROI; this ROI was repeatedly drawn on three adjacent slices, and the mean

value was determined for each inserted plug-in material. The size of the ROI and measured position were replicated for all images that were used to calculate mean HU. The image noise values were determined from the standard deviations obtained from the ROIs. The SNR values were computed using the HU and noise values. SNR = HUspecified material/noise. For qualitative analysis, a visual evaluation of phantom images obtained from 4 different acquisition protocols was also performed by the same radiologists using a 3-point scale. The 3-point scale was formulated based on the appearance of streak artifacts on images and the image distortion. Point 3: artifacts and image distortion is absent, 2: medium artifacts and less image distortion, 1: higher artifacts and image distortion.

Results Phantom dose and image quality In phantom study, protocols-2 and -3 generated significantly lower mean dose values (PED: 30%; DAP: 29%) compared to protocol-1 (244 ± 7 mGy (242–246); 71.8 ± 4 Gy.cm2 (69–74)). Significantly lower doses were (33% and 34%) noted for protocol-4 (114 ± 8 mGy (112–116); 33.6 ± 3 Gy.cm2 (32–36)) compared to 3; however, image quality was seemingly poor with -4 (Fig 2A, B). Both protocols-1 and -4 were avoided for patient examinations due to a significant radiation dose (Fig 3A, B) and a poor

Table 5 Details of the image quality visual evaluation scores obtained from the evaluated radiologists. Results suggest protocol 1 provide a high image quality and this followed by protocol 2 and -3. Results of the image quality visual evaluation using 3-point scale on phantom images (5 scanned phantom image datasets) Points

Protocol 1

Protocol 2

Protocol 3

Protocol 4

1 2 3

0 0 5

0 5 0

0 5 0

5 0 0

Fig. 3. The radiation dose such as patient entrance dose (PED) in mGy and dose-area-product (DAP) in Gy cm2 calculated for four different CBCT acquisition protocols (P1, P2, P3 and P4) in patients as well as phantom. A significant reductions of dose can be observed from protocol-1 to -4 (p < 0.05); however, the difference of dose between protocol2 and -3 was insignificant (p > 0.05).

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

6

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

image quality (Table 5). Visual image quality evaluation using 3point scale, protocols-2 and -3 produced medium artifacts with less image distortion.

Patient dose and image quality Mean patient dose (Fig 3A, B) showed statistically insignificant difference (p > 0.05) between protocols-2 (PED: 121.7 ± 19 mGy (95–138); DAP: 32.6 ± 5.1 Gy.cm2 (27–38)) and -3 (PED: 119.9 ± 16 mGy (84–132); DAP: 31.9 ± 4.3 Gy.cm2 (24–34)). Phantom study manifests similar results (p > 0.05) as patient examinations using protocols-2 (170 ± 7 mGy (158–179); 50.8 ± 4 Gy.cm2 (46–54)) and -3 (169.2 ± 7 (161–175); 50.6 ± 3 Gy.cm2 (46–54)) regarding dose. The association between image quality (SNR) as a function of radiation dose for phantom (Fig 4A) and patients (Fig 4B) showed the SNR is strongly correlated (all R2 = 1) with PED.

Image quality and lesion delineation (Quantitative)

Fig. 4. Displays the correlation between signal-to-noise ratio and patient-entrancedose for phantom (A) and patient (B) investigations. The coefficient values show a good correlation between the image quality and dose for all tumor types and anatomical structures examined.

The HU and image noise estimated for the patient hepatic tumor and anatomical organs were shown in Fig 2C and D. Patient images showed strong correlations (all R2 > 0.95) exists between protocol-2 and -3 for image quality parameters (Fig 5). Mean HU values were insignificant (all p > 0.05) between protocols-2 and 3 for tumor categories and anatomical organs examined. Determined SNR shows a very strong correlation (R2 > 0.95) between protocols-2 and -3 for all tumor categories; the comparison between the protocols produced insignificant results (all p > 0.05). TLC values were insignificant (all p > 0.05) between protocols-2 and -3; in addition to that TLC was strongly correlated between these two protocols. Regarding lesion delineation, the HU, SNR, and CNR were insignificant (all p > 0.05) between protocols-2 and -3 (Fig. 6A–C) during comparison.

Fig. 5. Correlation between two image acquisition patient protocols 2 and -3. The patient image quality parameters such as Hounsfield Unit (A), image noise (B), signal-tonoise ratio (C) and lesion conspicuity determinant TLC values (D) shows strong correlation between these two protocols. The generated correlation coefficient (R2) for all image quality parameters are displayed within the associate graphs.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

Image quality and lesion delineation (Qualitative) Qualitative lesion delineation generated insignificant (all p > 0.05) results between protocols-2 and -3 in all lesion categories (Fig 7). Inter-reader comparison showed almost perfect agreement (k = 0.81–0.99) between scores obtained from the three readers.

7

higher radiation dose and protocol-4 used the lowest doses for imaging of the phantom. This fewer dose of protocol-4 generated distorted images with higher artifacts, which was not useful for diagnostic purposes. Ethical considerations prevented patient examinations using all presented CBCT protocols especially

Lesion detectability and sensitivity Lesion sensitivity and predictive values on CBCT were determined using MR images for the two examined acquisition protocols (Table 6). Computed values manifested very close lesion sensitivity and predictive values for both CBCT protocols-2 and 3. The CBCT image data produced 81–89% sensitivity during protocol-2 and 81–85% during protocol-3 examination. The predictive values ranging from 94.6 to 95.6% with protocol-2 and 92– 94.6% with protocol-3 were obtained for three tumor categories. Discussion CBCT can be used for the imaging of patients during various real-time image-guided interventional therapeutic procedures [15–17]. Generation of scatter radiation, artifacts and image distortion are some of the limiting factors affecting image quality, thereby causing increased radiation dose to the examined patients [7,18,19]. Present study evaluated patient dose from CBCT imaging during image-guided transarterial chemoembolization procedures using different image acquisition protocols. The electron density phantom combined with CTDI body phantom was used to simulate the patient body during phantom examination. Both protocols-2 and -3 delivered similar patient dose during CBCT examinations as these two acquisition protocols used the same frame numbers (248 frames) to generate image datasets. The protocol-1 used a

Fig. 6. Cone beam CT images generated using acquisition protocols-2 (A) and -3 (B) during hepatic transarterial chemoembolization. White arrow shows the hepatic lesion, and white circle represent the ROIs used to calculate HU of the liver parenchyma. Box-plot shows the contrast-to-noise ratio (CNR) distribution (C) for different hepatic lesions such as homogenously enhanced lesion (HOEL), heterogeneously enhanced lesion (HEEL) and non-enhanced lesion (NEL).

Fig. 7. Lesion delineation qualitative assessment scores obtained from the three readers for different lesions types (Homogenously enhanced lesion- HOEL, Heterogeneously enhanced lesion- HEEL, and Unenhanced lesion- UEL). P2 and P3 are the patient imaging protocols and the comparison between these protocols revealed insignificant p-values.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

8

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx

Table 6 Shows calculated lesion sensitivity and predictive values using MRI and cone-beam CT image data acquisition protocols 2 and -3. (HOEL: Homogenously enhanced lesion, HEEL: Heterogeneously enhanced lesion, UEL: Unenhanced lesion). Present study included the hepatic lesions P1 cm diameter size to calculate sensitivity and predictive values. Lesion type

HOEL HEEL UEL HOEL HEEL UEL

Lesion on image data MR-image

CBCT Protocol 2

65 53 61 MR- image 58 43 41

61 45 55 CBCT Protocol 3 51 37 38

True positive

False negative

False positive

True negative

Sensitivity (%)

Specificity (%)

Positive predictive value (%)

Negative predictive value (%)

58 43 52

7 10 9

3 2 3

0 0 0

89 81 85.3

0 0 0

95 95.6 94.6

0 0 0

47 35 35

11 8 6

4 2 3

0 0 0

81 81.4 85

0 0 0

92.2 94.6 92

0 0 0

acquisition protocol with a higher number of frames that delivered higher patient dose, and fewer frame numbers which considerably reduced image quality as well as lesion delineation on images. Due to a higher dose and non-diagnostic image quality during phantom study, the protocols-1 and -4 were eliminated from the patient image acquisition protocols. Protocol-1 used a higher number of frames and higher rotation angle to generate image data; these parameters contributed a higher radiation dose to the phantom. Reduction of dose related to protocol-4 was due to the reduction of the number of frames from 397 to 166, and the angle covered reduced from 360° to 180°. A previous publication manifests that CBCT imaging with 5 s acquisition protocol was used for comparison of different contrast material effect on image quality [2]. They arrived at a patient dose of 111.7 mGy and 29.2Gycm2 during single tube-detector rotation using 300 frames; furthermore, these values are very close to the dose values generated in protocol-2 in this study. Regarding image quality, a direct comparison is impossible with the present study due to the use of double exposure generating image data in the reference. There are recent publications discussing dose, image quality and advantages of CBCT in radiation therapy (RT). The role of currently using CBCT dose indicators in RT were evaluated by Rampado et al. [20], and these indicators were correlated with effective and organ dose values. The phantom radiation dose indicators showed identical information relate to organ dose measured for a standard patient. Foley et al. [21] showed on-board CBCT image registration feasibility to planning CT using a 3D phasecorrelation algorithm, which offered benefits to reduce image noise levels on images, and to adapt the target dose on a fraction by fraction basis in RT. Similarly, a phase correlation technique was used by Wang et al. [22] to register 2D portal images in RT, and a 3D registration of MR images of brain was investigated by Kojima et al. [23]. In another study, a partial angle CBCT imaging with three different paths were investigated for head and neck region to evaluate image quality and absorbed dose to the critical organs [24]. The scanning direction showed negligible variations on image quality; however, the doses on critical organs were significantly different in all paths. The aim of optimization process in patient imaging is to determine the lowest possible dose level to generate an optimal image quality needed for diagnosis and/or therapeutic procedure. Patient image quality parameters manifested good results with insignificant differences between acquisition protocols-2 and-3 in this study. The generated SNR and CNR values displayed statistically insignificant differences between these protocols with an acceptable patient dose. The determined TLC values showed that both protocols are good for lesion delineation; again, these protocols generated similar or closer values for a certain lesion category. Qualitative analysis also showed similar results due to the same number of acquisition frames used to generate cross-sectional image data in both protocols. Lesion detection, sensitivity and

predictive values were favorable for both examined patient protocols. Hence, we believe that imaging protocol-2 and -3 are beneficial for patients compared to other CBCT protocols examined due to a good image quality, lesion detectability, and delineation with acceptable patient dose.

Conclusion CBCT image acquisition performed using protocols-2 and -3 produced good image quality for clinical image analysis with acceptable patient dose during image-guided procedures. Good lesion detectability and delineation were achieved using these patient acquisition protocols with CM injection during different hepatic lesion analysis. Patient dose during CBCT imaging was dependent on the acquired number of frames, frame rate, tubedetector rotation angle, applied tube potential, tube current, beam filtration and lesion contrast material uptake. A further reduction of frame numbers certainly reduces patient dose and increases patient radiation safety, but this reduction results in a reduction in image quality.

References [1] Paul J, Jacobi V, Farhang M, et al. Radiation dose and image quality of X-ray volume imaging systems: cone beam computed tomography, digital subtraction angiography, and digital fluoroscopy. Eur Radiol 2013;23 (6):1582–93. [2] Paul J, Vogl TJ, Mbalisike EC. Radiation dose and image quality evaluation relative to different contrast media using cone beam CT. Imag Med 2012;4:505–13. [3] Baba R, Konno Y, Ueda K, et al. Comparison of flat-panel detector and imageintensifier detector for cone-beam CT. Comput Med Imaging Graph 2002;26:153–8. [4] Dijkstra ML, Eagleton MJ, Greenberg RK, et al. Intraoperative C-arm cone- beam computed tomography in fenestrated/branched aortic endografting. J Vasc Surg 2011;53(3):583–90. [5] Jin KN, Park CM, Goo JM, et al. Initial experience of percutaneous transthoracic needle biopsy of lung nodules using C-arm cone- beam CT systems. Eur Radiol 2010;20:2108–15. [6] Ishikura R, Ando K, Nagami Y, et al. Evaluation of vascular supply with conebeam computed tomography during intra-arterial chemotherapy for a skull base tumor. Radiat Med 2006;24:384–7. [7] Siewerdsen JH, Jaffray DA. Cone-beam computed tomography with a flat-panel imager: magnitude and effects of x-ray scatter. Med Phys 2001;28:220–31. [8] Endo M, Mori S, Tsunoo T, et al. Magnitude and effects of x-ray scatter in a 256slice CT scanner. Med Phys 2006;33:3359–68. [9] Lee H, Xing L, Lee R, et al. Scatter correction in cone-beam CT via a half beam blocker technique allowing simultaneous acquisition of scatter and image information. Med Phys 2012;39:2386–95. [10] las Heras de H, Rampado O, Andersson J, et al. Quality controls for CBCT devices: The EFOMP guideline for quality assurance of images and dose. http:// dx.doi.org/10.1016/j.ejmp.2016.07.626. [11] http://www.sunnuclear.com/documents/datasheets/gammex/ct_electron_ density_phantom.pdf. [12] Paul J, Mbalisike EC, Vogl TJ. Radiation dose to procedural personnel and patients from an X-ray volume imaging system. Eur Radiol 2013;23 (12):3262–70.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111

J. Paul et al. / Physica Medica xxx (2016) xxx–xxx [13] Sakamoto H, Aikawa Y, Ikegawa H, et al. Consideration of the newly standardized interventional reference point. Nihon Hoshasen Gijutsu Gakkai Zasshi 2004;60:520–7. [14] Koelblinger C, Schima W, Berger-Kulemann V, et al. C-arm CT during hepatic arteriography tumour-to-liver contrast: intraindividual comparison of three different contrast media application protocols. Eur Radiol 2013;23:938–42. [15] Miyayama S, Yamashiro M, Hattori Y, et al. Efficacy of cone-beam computed tomography during transcatheter arterial chemoembolization for hepatocellular carcinoma. Jpn J Radiol 2011;29:371–7. [16] Xu J, Reh DD, Carey JP, et al. Technical assessment of a cone-beam CT scanner for otolaryngology imaging: image quality, dose, and technique protocols. Med Phys 2012;39(8):4932–42. [17] Yu L, Vrieze TJ, Bruesewitz MR, et al. Dose and image quality evaluation of a dedicated cone beam CT system for high contrast neurologic applications. AJR Am J Roentgenol 2010;194(2):W193–201. [18] Elstrom UV, Muren LP, Petersen JBB, et al. Evaluation of image quality for different kV cone-beam CT acquisition and reconstruction methods in the head and neck region. Acta Oncol 2011;50:908–17.

9

[19] Orth RC, Wallace MJ, Kuo MD. C-arm cone-beam CT: general principles and technical considerations for use in interventional radiology. J Vasc Interv Radiol 2008;19:814–21. [20] Rampado O, Giglioli FR, Rossetti V, et al. CBCT dose indicators: Evaluation of different approaches and correlation with patient dose. http://dx.doi.org/10. 1016/j.ejmp.2016.01.305. [21] Foley D, O’Brien DJ, Luis León-Vintró, et al. Phase correlation applied to the 3D registration of CT and CBCT image volumes. http://dx.doi.org/10.1016/j.ejmp. 2016.02.009. [22] Wang JZ, Reinstein LE. Investigation of a phase-only correlation technique for anatomical alignment of portal images in radiation therapy. Phys Med Biol 1996;41:1045–58. [23] Kojima S, Hirata M, Shinohara H, Ueno E. Reproducibility of scan prescription in follow-up brain MRI: manual versus automatic determination. Radiol Phys Technol 2013:375–84. [24] Khamfongkhruea C, Utapom K, Munsing S, et al. Posterior kV-CBCT scanning of the head and neck region minimizes doses to critical organs with sustained image quality. http://dx.doi.org/10.1016/j.ejmp.2015.03.016.

Please cite this article in press as: Paul J et al. Effect of diagnostic cone-beam computed tomography protocols on image quality, patient dose, and lesion detection. Phys. Med. (2016), http://dx.doi.org/10.1016/j.ejmp.2016.11.111