Deep learning for detection and 3D segmentation of maxillofacial bone lesions in cone beam CT.

OBJECTIVES: To develop an automated deep-learning algorithm for detection and 3D segmentation of incidental bone lesions in maxillofacial CBCT scans. METHODS: The dataset included 82 cone beam CT (CBCT) scans, 41 with histologically confirmed benign bone lesions (BL) and 41 control scans (without lesions), obtained using three CBCT devices with diverse imaging protocols. Lesions were marked in all axial slices by experienced maxillofacial radiologists. All cases were divided into sub-datasets: training (20,214 axial images), validation (4530 axial images), and testing (6795 axial images). A Mask-RCNN algorithm segmented the bone lesions in each axial slice. Analysis of sequential slices was used for improving the Mask-RCNN performance and classifying each CBCT scan as containing bone lesions or not. Finally, the algorithm generated 3D segmentations of the lesions and calculated their volumes. RESULTS: The algorithm correctly classified all CBCT cases as containing bone lesions or not, with an accuracy of 100%. The algorithm detected the bone lesion in axial images with high sensitivity (95.9%) and high precision (98.9%) with an average dice coefficient of 83.5%. CONCLUSIONS: The developed algorithm detected and segmented bone lesions in CBCT scans with high accuracy and may serve as a computerized tool for detecting incidental bone lesions in CBCT imaging. CLINICAL RELEVANCE: Our novel deep-learning algorithm detects incidental hypodense bone lesions in cone beam CT scans, using various imaging devices and protocols. This algorithm may reduce patients' morbidity and mortality, particularly since currently, cone beam CT interpretation is not always preformed. KEY POINTS: • A deep learning algorithm was developed for automatic detection and 3D segmentation of various maxillofacial bone lesions in CBCT scans, irrespective of the CBCT device or the scanning protocol. • The developed algorithm can detect incidental jaw lesions with high accuracy, generates a 3D segmentation of the lesion, and calculates the lesion volume.

Feasibility and Accuracy of a Novel Hands-Free Robotic System for Percutaneous Needle Insertion and Steering.

PURPOSE: To assess the performance and accuracy of CT-guided needle insertion for clinical biopsies using a novel, hands-free robotic system that balances accuracy with the duration of the procedure and radiation dose. MATERIALS AND METHODS: A prospective, multi-center study was conducted on 60 clinically indicated biopsies of abdominal lesions at two centers (Center 1, n=26; Center 2, n=34). CT datasets were obtained for planning and controlled placement of 17g and 18g needles using a patient-mounted, CT-guided robotic system with 5 degrees of freedom. Planning included target selection, skin entry point, and predetermined checkpoints where additional imaging was performed to permit stepwise correction of the needle trajectory. Success rate, needle tip-to-target distance, number of checkpoints used, number of trajectory corrections, procedure duration, and effective radiation dose were recorded and compared between centers. RESULTS: In 55 of 60 procedures (91.7%), the robot positioned the trocar needle successfully on target. In the remaining 5 patients, the procedure was manually performed by the operator due to technical failure (n=3) or patient-related factors (n=2). The average lesion size was 2.8 ± 1.7cm with a lesion depth from the skin of 8.7 ± 2.6cm, and there was no difference between centers. The overall accuracy (needle tip-to-target distance) was 1.71 ± 1.49 (range 0.05-7.20mm), with an accuracy of 2.06 ± 1.45 mm at Center 1 and 1.45 ± 1.52 mm at Center 2 (p=0.1358). Center 1 used significantly more checkpoints (4.96 ± 1.08) and performed target adjustments in 20 of 24 (83%) cases compared to Center 2 (2.77 ± 0.6 checkpoints and target adjustments in 13 of 31 cases, 42%) (p=0.0024). Accordingly, the steering duration from skin entry to the target varied between Centers 1 and 2; 13.1min ± 4.25min vs. 5.7min ± 2.7min, respectively (p <0.001). The average DLP for the entire procedure was 1147 ± 820 mGycm, with a slightly lower average at Center 2 (1031 ± 724 mGycm) compared to Center 1 (1297 ± 925 mGycm) (p=0.236). CONCLUSION: Accurate needle-targeting within an error of 2mm can be achieved in patients using a CT-guided robotic system. The variation in the number of checkpoints did not affect system accuracy but was related to shorter steering times and may contribute to a lower radiation dose. Accurate needle insertion using a hands-free CT-guided robotic system may facilitate difficult needle placement and enhance the performance of less-experienced interventionalists.

Spectral material characterization with dual-energy CT: comparison of commercial and investigative technologies in phantoms.

BACKGROUND: Dual-energy computed tomography (DECT) enables tissue discrimination based on the X-ray attenuations at different photon energies emitted by the tube. The spectral dependencies of net X-ray attenuation can be analyzed and used to characterize specific materials. PURPOSE: To evaluate the capability of DECT to characterize and differentiate high-density materials, using spectral analysis. MATERIAL AND METHODS: Images of phantoms containing iodine, barium, gadolinium, and calcium solutions in five concentrations were obtained from three DECT scanners and with sequential scanning at different kV values from three conventional MDCT devices. DECT studies were performed with commercial dual-source and rapid kV-switching systems, and a spectral-detector CT (SDCT) prototype based on dual-layer detector technology. Spectral maps describing Hounsfield Units (HU) in low- versus high-energy images were calculated and characterizing curves for all materials were compared. RESULTS: Spectral low- to- high energy maps yielded linear curves (R(2) = 0.98-0.999) with increasing slopes for calcium, gadolinium, barium, and iodine, respectively. Slope differences between all material pairs were highest (reaching 45%) for DECT with dual-source (140/80 kV) and rapid kV-switching (60/80 keV), reaching statistical significance (P < 0.05) with most techniques. Slope differences between all material pairs for sequential scanning were lower (reaching 32%). Slope differences lacked statistical significance for iodine-barium with two sequential-acquisition techniques and the dual-source DECT scanner, and the calcium-gadolium pair with the dual-source scanner. CONCLUSION: All designated techniques for dual-energy scanning provide robust and material-specific spectral characterization and differentiation of barium, iodine, calcium, and gadolinium, though to varying degrees.

Evaluation of the Pulmonary Arteries on CTPA With Dual Energy CT: Objective Analysis and Subjective Preferences in a Multireader Study.

PURPOSE: To perform qualitative and quantitative evaluation of low-monoenergetic images (50 KeV) compared with conventional images (120 kVp) in pulmonary embolism (PE) studies and to determine the extent and clinical relevance of these differences as well as radiologists' preferences. MATERIALS AND METHODS: One hundred fifty CT examinations for PE detection conducted on a single-source dual-energy CT were retrospectively evaluated. Attenuation, contrast-to-noise-ratio, and signal-to-noise-ratio were obtained in a total of 8 individual pulmonary arteries on each exam-including both central (450/1200=37.5%) and peripheral (750/1200=62.5%) locations. Results were compared between the conventional and low-monoenergetic images. For quality assessment, 41 images containing PE were presented side-by-side as pairs of slices in both conventional and monoenergetic modes and evaluated for ease in embolus detection by 9 radiologists: cardiothoracic specialists (3), noncardiothoracic specialists (3), and residents (3). Paired samples t tests, a-parametric Wilcoxon test, McNemar test, and kappa statistics were performed. RESULTS: Monoenergetic images had an overall statistically significant increased average ratio of 2.09 to 2.26 ( P <0.05) for each measured vessel attenuation, with an increase in signal-to-noise ratio (23.82±9.29 vs. 11.39±3.2) and contrast-to-noise ratio (17.17±6.7 vs 7.27±2.52) ( P <0.05). Moreover, 10/150 (6%) of central pulmonary artery measurements considered suboptimal on conventional mode were considered diagnostic on the monoenergetic images (181±14.6 vs. 387.7±72.4 HU respectively, P <0.05). In the subjective evaluation, noncardiothoracic radiologists showed a preference towards low-monoenergetic images, whereas cardiothoracic radiologists did not (74.4% vs. 57.7%, respectively, P <0.05). CONCLUSIONS: The SNR and CNR increase on monoenergetic images may have clinical significance particularly in the setting of sub-optimal PE studies. Noncardiothoracic radiologists and residents prefer low monoenergetic images.

Patient radiation exposure during percutaneous endovascular revascularization of the lower extremity.

BACKGROUND: Percutaneous endovascular revascularization is emerging as the first line treatment for peripheral artery disease for both intermittent claudication and chronic critical limb ischemia. Radiation doses for these interventions have not been well documented. METHODS: A single center retrospective study of therapeutic endovascular lower extremity interventions performed between September 2006 and December 2011 was undertaken. Collected data included patient demographics, procedure indication, procedural access, anatomic location of occlusive disease (pelvis, thigh, below-knee, or multilevel), and radiation exposure parameters including dose area product (DAP) and fluoroscopy time. RESULTS: Data was available for 382 procedures performed in 313 patients. Eighteen procedures bilateral procedures were excluded. Access site and complete anatomic data were available for 346 procedures. DAP was significantly higher for procedures performed in the pelvis compared with thigh procedures (179.6 vs 63.2 Gy*cm(2); P < .0001) and below-knee procedures (179.6 vs 28.9 Gy*cm(2); P < .0001), despite shorter fluoroscopy times (11.8 vs 16.4 minutes; P < .0001 and 11.1 vs 31.06 minutes; P < .0001, respectively). Procedure access-site affected radiation dose as well; contralateral up-and-over access resulted in a higher DAP than antegrade access (112.2 vs 42.6 Gy*cm(2); P < .0001). In a multivariable analysis, anatomic location of the procedure showed the strongest association with radiation dose (P < .0001). CONCLUSIONS: Percutaneous endovascular revascularization for lower extremity peripheral artery disease involves a substantial radiation dose, comparable, on average, to a computed tomography scan of the abdomen and pelvis. Procedures performed in the pelvis for intermittent claudication involve more radiation than thigh or below-knee procedures for chronic critical limb ischemia. Radiation dose should be considered when planning these procedures.

Quantitative assessment of renal obstruction in multi-phase CTU using automatic 3D segmentation of the renal parenchyma and renal pelvis: A proof of concept.

Purpose: Quantitative evaluation of renal obstruction is crucial for preventing renal atrophy. This study presents a novel method for diagnosing renal obstruction by automatically extracting objective indicators from routine multi-phase CT Urography (CTU). Material and methods: The study included multi-phase CTU examinations of 6 hydronephrotic kidneys and 24 non-hydronephrotic kidneys (23,164 slices). The developed algorithm segmented the renal parenchyma and the renal pelvis of each kidney in each CTU slice. Following a 3D reconstruction of the parenchyma and renal pelvis, the algorithm evaluated the amount of the contrast media in both components in each phase. Finally, the algorithm evaluated two indicators for assessing renal obstruction: the change in the total amount of contrast media in both components during the CTU phases, and the drainage time, "T1/2", from the renal parenchyma. Results: The algorithm segmented the parenchyma and renal pelvis with an average dice coefficient of 0.97 and 0.92 respectively. In all the hydronephrotic kidneys the total amount of contrast media did not decrease during the CTU examination and the T1/2 value was longer than 20 min. Both indicators yielded a statistically significant difference (p < 0.001) between hydronephrotic and normal kidneys, and combining both indicators yielded 100% accuracy. Conclusions: The novel algorithm enables accurate 3D segmentation of the renal parenchyma and pelvis and estimates the amount of contrast media in multi-phase CTU examinations. This serves as a proof-of-concept for the ability to extract from routine CTU indicators that alert to the presence of renal obstruction and estimate its severity.

A novel method for estimating the urine drainage time from the renal collecting system.

PURPOSE: Partial obstruction of the upper urinary tract is a common urological pathology that leads to progressive atrophy and dysfunction of the kidney. Most methods for evaluating the urine drainage rate, to assess the severity of partial obstruction, involve injection of markers into the blood stream and therefore the filtration rate from the blood effects the drainage rate. This study presents a novel method for assessing the drainage rate from the upper urinary tract by analyzing sequential fluoroscopic images from a routine nephrostogram, in which contrast material is introduced directly into the renal collecting system. METHODS: Fluoroscopic images from 36 nephrostograms, following percutaneous nephrolithotomy, were retrospectively evaluated, 19 with a dilated renal pelvis. A radiological model for calculating the radiopacity of the renal pelvis, which reflects the amount of contrast material in each sequential image, was developed. Using this model, an algorithm was designed for generating a drainage curve and calculating the "drainage time" t1/2 in which half of the contrast material has drained from the renal pelvis. RESULTS: Analysis of images of a step-wedge phantom made of an increasing number of contrast material layers showed that the calculated radiopacity of each step was proportional to the amount of contrast material, independent of the background attenuation. Analysis of the nephrostograms showed that the drainage curves highly fitted an exponential function (R = 0.961), with a significantly higher t1/2 for dilated cases. CONCLUSION: The developed method may be used for a quantitative and accurate estimation of the urine drainage rate.

Clinical evaluation of a robotic system for precise CT-guided percutaneous procedures.

PURPOSE: To assess accuracy and compare protocols for CT-guided needle insertion for clinical biopsies using a hands-free robotic system, balancing system accuracy with duration of procedure and radiation dose. METHODS: Thirty-two percutaneous abdominal and pelvic biopsies were performed and analyzed at two centers (Center 1 n = 11; Center 2 n = 21) as part of an ongoing prospective, multi-center study. CT datasets were obtained for planning and controlled placement of 17 g needles using a patient-mounted, CT-guided robotic system. Planning included target selection, skin entry point, and predetermined checkpoints. Additional CT imaging was performed at checkpoints to confirm needle location and permit stepwise correction of the trajectory. Center 1 used a more conservative approach with multiple checkpoints, whereas Center 2 used fewer checkpoints. Scanning and needle advancement were performed under respiratory gating. Accuracy, radiation dose, and steering duration were compared. RESULTS: Overall accuracy was 1.6 ± 1.5 mm (1.9 ± 1.2 mm Center 1; 1.5 ± 1.6 mm Center 2; p = 0.55). Mean distance to target was 86.2 ± 27.1 mm (p = 0.18 between centers). Center 1 used 4.6 ± 0.8 checkpoints, whereas Center 2 used 1.8 ± 0.6 checkpoints (p < 0.001). Effective radiation doses were lower for Center 1 than for Center 2 (22.2 ± 12.6 mSv vs. 11.7 ± 4.3 mSv; p = 0.002). Likewise, steering duration (from planning to target) was significantly reduced in relation to the number of checkpoints from 43.8 ± 15.9 min for Center 1 to 30.5 ± 10.2 min for Center 2 (p = 0.008). CONCLUSIONS: Accurate needle targeting with < 2 mm error can be achieved in patients when using a CT-guided robotic system. Judicious selection of the number of checkpoints may substantially reduce procedure time and radiation dose without sacrificing accuracy.

Dual energy CT in acute appendicitis: value of low mono-energy.

OBJECTIVES: To assess the potential role of low monoenergetic images in the evaluation of acute appendicitis. METHODS: A retrospective study of 42 patients with pathology proven acute appendicitis underwent contrast-enhanced-CT conducted on a single-source-DECT before surgery. Attenuation, SNR, and CNR were calculated on both monoenergetic and conventional images and compared to 24 abdominal CT-scans with normal appendix. Representative conventional and monoenergetic images were randomized and presented side-by-side to three abdominal radiologists to determine preferred images for detecting inflammation. Additionally, six individual acute inflammatory characteristics were graded on a 1-5 scale to determine factors contributing to differences between conventional and monoenergetic images by 2 abdominal radiologists. Paired t-tests, Wilcoxon and McNemar tests, and intra-observer error statistics were performed. RESULTS: For the inflamed appendixes monoenergetic images had overall increased attenuation (average ratio 1.7; P < 0.05), signal-to-noise-ratio (6.7 ± 3.1 vs 4.2 ± 1.6; P < 0.001) and contrast-to-noise-ratio (12.1 ± 3 vs 9 ± 2.1; P < 0.001). Moreover, this increase was not found in normal appendixes (P < 0.001 vs p = 0.28-0.44). Subjectively, radiologists showed significant preferences towards monoenergetic images (P < 0.001), with inter-reader agreement of 0.84. Two parameters, diffuse bowel wall and mucosal enhancement, received significantly higher scores on monoenergetic images (average 4.3 vs. 3.0; P < 0.001 and 2.8 vs. 2.3 P < 0.03 respectively, with interobserver agreements of 62% and 52%). CONCLUSION: Increased bowel wall conspicuity from enhanced attenuation, SNR, and CNR on low monenergetic CT images results in a significant preference by radiologists for these images when assessing acute inflamed appendixes. Thus, close inspection of low monoenergetic images may improve the visualization of acute inflammatory bowel processes.

An artificial intelligence system using machine-learning for automatic detection and classification of dental restorations in panoramic radiography.

OBJECTIVES: The aim of this study was to develop a computer vision algorithm based on artificial intelligence, designed to automatically detect and classify various dental restorations on panoramic radiographs. STUDY DESIGN: A total of 738 dental restorations in 83 anonymized panoramic images were analyzed. Images were automatically cropped to obtain the region of interest containing maxillary and mandibular alveolar ridges. Subsequently, the restorations were segmented by using a local adaptive threshold. The segmented restorations were classified into 11 categories, and the algorithm was trained to classify them. Numerical features based on the shape and distribution of gray level values extracted by the algorithm were used for classifying the restorations into different categories. Finally, a Cubic Support Vector Machine algorithm with Error-Correcting Output Codes was used with a cross-validation approach for the multiclass classification of the restorations according to these features. RESULTS: The algorithm detected 94.6% of the restorations. Classification eliminated all erroneous marks, and ultimately, 90.5% of the restorations were marked on the image. The overall accuracy of the classification stage in discriminating between the true restoration categories was 93.6%. CONCLUSIONS: This machine-learning algorithm demonstrated excellent performance in detecting and classifying dental restorations on panoramic images.

Can the size of microcalcifications predict malignancy of clusters at mammography?

RATIONALE AND OBJECTIVES: The purpose of this study was to determine whether the size of mammographically detected microcalcifications is predictive of malignancy. MATERIALS AND METHODS: Two hundred sixty mammograms showing clustered microcalcifications with proven diagnoses (160 malignant, 100 benign) were respectively reviewed by experienced mammographers. Lesions that were obviously benign in appearance were excluded from the study. A computer-aided diagnosis system digitized the lesions at 600 dpi, and the microcalcifications on the digital image were interactively defined by mammographers. Subsequently, three quantitative features that reflected the size of the microcalcifications-length, area, and brightness-were automatically extracted by the system. For each feature, the standard average of values obtained for individual calcifications within the cluster and the average with emphasis on extreme values (E) obtained in a single cluster were analyzed and matched with pathologic results. RESULTS: In the malignant group of cases, the mean values of the standard average length and area were significantly higher (P < .0001) than the mean values in the benign group. Distribution analysis demonstrated that an average length of more than 0.41 mm was associated with malignant lesions 77% of the time, while an average length of less than 0.41 mm was associated with benign lesions 71% of the time. The mean of the average length (E) and area (E) of microcalcifications within the cluster demonstrated an even higher discriminative power when compared with the standard average length and area. The average brightness, on the other hand, showed only a low discriminative power. CONCLUSION: Digital computerized analysis of mammographically detected calcifications demonstrated that the average length and area of the calcifications in benign clusters were significantly smaller than those in malignant clusters.

Significance of enhanced cerebral gray-white matter contrast at 80 kVp compared to conventional 120 kVp CT scan in the evaluation of acute stroke.

We aimed to determine whether 80 kVp conventional nonenhanced head CT scans have better gray-white matter contrast than standard 120 kVp scans performed on the same patients. Thirty head CT scans acquired at 80 kVp (CT dose index [CTDI]vol 46) were compared to prior studies in the same patients performed at 120 kVp (CTDIvol 59). Signal (Hounsfield units [HU]), noise (sd HU), and contrast-to-noise ratio per dose (CNRD) were assessed in multiple cerebral gray and white matter regions of interest. A noise correction factor was used to compensate for scanning at different CTDIvol values. Average gray matter signal at 80 kVp and 120 kVP was 33.9 ± 3.5 HU and 29 ± 4.6 HU, respectively (p<0.0001); the averages for white matter were 22.5 ± 3.1 HU and 21.6 ± 4.6 HU, respectively (p=0.11). Corrected noise was 3 ± 0.6 and 2.7 ± 0.6, respectively, for gray matter (p=0.0001), and 2.8 ± 0.6 and 2.6 ± 0.5, respectively, for white matter (p=0.00001). The gray-white matter CNRD was 4.0 ± 1.2 at 80 kVp and 2.8 ± 1 at 120 kVp (p<0.00001). Cerebral gray-white matter CNRD is increased by 40% at 80 kVp compared to conventional 120 kVp CT scans. These findings justify further clinical evaluation in the acute stroke setting.

Stratification of mammographic computerized analysis by BI-RADS categories.

The Breast Imaging Reporting and Data System (BI-RADS) was implemented to standardize characterization of mammographic findings. The purpose of the present study was to evaluate in which BI-RADS categories the changes recommended by computerized mammographic analysis are most beneficial. Archival cases including, 170 masses (101 malignant, 69 benign) and 63 clusters of microcalcifications (MCs; 36 malignant, 27 benign), were evaluated retrospectively, using the BI-RADS categories, by several radiologists, blinded to the pathology results. A computerized system then automatically extracted from the digitized mammogram features characterizing mammographic lesions, which were used to classify the lesions. The results of the computerized classification scheme were compared, by receiver operating characteristics (ROC) analysis, to the conventional interpretation. In the "low probability of malignancy group" (excluding BI-RADS categories 4 and 5), computerized analysis improved the A(z )of the ROC curve significantly, from 0.57 to 0.89. In the "high probability of malignancy group" (mostly category 5) the computerized analysis yielded an ROC curve with an A(z )of 0.99. In the "intermediate probability of malignancy group" computerized analysis improved the A(z )significantly, from 0.66 for to 0.83. Pair-wise analysis showed that in the latter group the modifications resulting from computerized analysis were correct in 83% of cases. Computerized analysis has the ability to improve the performance of the radiologists exactly in the BI-RADS categories with the greatest difficulties in arriving at a correct diagnosis. It increased the performance significantly in the problematic group of "intermediate probability of malignancy" and pinpointed all the cases with missed cancers in the "low probability" group.

Computer-aided classification of BI-RADS category 3 breast lesions.

PURPOSE: To evaluate a system for computer-aided classification (CAC) of lesions assigned to Breast Imaging Reporting and Data System (BI-RADS) category 3 at conventional mammographic interpretation. MATERIALS AND METHODS: A CAC system was used to analyze 106 cases of lesions (42 malignant) that at blinded retrospective interpretation were assigned to BI-RADS category 3 by at least two of four radiologists. The CAC system automatically extracted from the digitized mammograms quantitative features that characterized the lesions. The system then used a classification scheme to score the lesions by the likelihood of their malignancy on the basis of these features. The classification scheme was trained with 646 pathologically proved cases (323 malignant), and the results were tested with receiver operating characteristic (ROC) analysis by using the jackknife method. Sensitivity, specificity, positive predictive value, and accuracy were calculated. Category 3 lesions were stratified among BI-RADS categories 2-5 according to CAC-assigned lesion score, and this classification was compared with the results of pathologic analysis. RESULTS: Jackknife analysis of CAC results in the training data set yielded a sensitivity of 94%, specificity of 78%, positive predictive value of 81%, and area under the ROC curve of 0.90. Of the 42 malignant lesions that had been classified at conventional interpretation as probably benign, nine were assigned by the CAC system to BI-RADS category 4, and 29 were assigned to category 5. The CAC system correctly upgraded the BI-RADS classification of these 38 lesions (sensitivity, 90%) and incorrectly upgraded the classification of only 20 benign lesions (specificity, 69%). CONCLUSION: The CAC system scored 38 of the 42 malignant lesions initially assigned to BI-RADS category 3 as BI-RADS category 4 or 5, and thus correctly upgraded the category in 90% of these lesions.