Evaluation of deep learning-based autosegmentation in breast cancer radiotherapy

Purpose To study the performance of a proposed deep learning-based autocontouring system in delineating organs at risk (OARs) in breast radiotherapy with a group of experts. Methods Eleven experts from two institutions delineated nine OARs in 10 cases of adjuvant radiotherapy after breast-conserving surgery. Autocontours were then provided to the experts for correction. Overall, 110 manual contours, 110 corrected autocontours, and 10 autocontours of each type of OAR were analyzed. The Dice similarity coefficient (DSC) and Hausdorff distance (HD) were used to compare the degree of agreement between the best manual contour (chosen by an independent expert committee) and each autocontour, corrected autocontour, and manual contour. Higher DSCs and lower HDs indicated a better geometric overlap. The amount of time reduction using the autocontouring system was examined. User satisfaction was evaluated using a survey. Results Manual contours, corrected autocontours, and autocontours had a similar accuracy in the average DSC value (0.88 vs. 0.90 vs. 0.90). The accuracy of autocontours ranked the second place, based on DSCs, and the first place, based on HDs among the manual contours. Interphysician variations among the experts were reduced in corrected autocontours, compared to variations in manual contours (DSC: 0.89–0.90 vs. 0.87–0.90; HD: 4.3–5.8 mm vs. 5.3–7.6 mm). Among the manual delineations, the breast contours had the largest variations, which improved most significantly with the autocontouring system. The total mean times for nine OARs were 37 min for manual contours and 6 min for corrected autocontours. The results of the survey revealed good user satisfaction. Conclusions The autocontouring system had a similar performance in OARs as that of the experts’ manual contouring. This system can be valuable in improving the quality of breast radiotherapy and reducing interphysician variability in clinical practice.

Balancing Speed and Accuracy in Cardiac Magnetic Resonance Function Post-Processing: Comparing 2 Levels of Automation in 3 Vendors to Manual Assessment

Automating cardiac function assessment on cardiac magnetic resonance short-axis cines is faster and more reproducible than manual contour-tracing; however, accurately tracing basal contours remains challenging. Three automated post-processing software packages (Level 1) were compared to manual assessment. Subsequently, automated basal tracings were manually adjusted using a standardized protocol combined with software package-specific relative-to-manual standard error correction (Level 2). All post-processing was performed in 65 healthy subjects. Manual contour-tracing was performed separately from Level 1 and 2 automated analysis. Automated measurements were considered accurate when the difference was equal or less than the maximum manual inter-observer disagreement percentage. Level 1 (2.1 ± 1.0 min) and Level 2 automated (5.2 ± 1.3 min) were faster and more reproducible than manual (21.1 ± 2.9 min) post-processing, the maximum inter-observer disagreement was 6%. Compared to manual, Level 1 automation had wide limits of agreement. The most reliable software package obtained more accurate measurements in Level 2 compared to Level 1 automation: left ventricular end-diastolic volume, 98% and 53%; ejection fraction, 98% and 60%; mass, 70% and 3%; right ventricular end-diastolic volume, 98% and 28%; ejection fraction, 80% and 40%, respectively. Level 1 automated cardiac function post-processing is fast and highly reproducible with varying accuracy. Level 2 automation balances speed and accuracy.

Characterization of Brain Stroke Using Image and Signal Processing Techniques

Cross-sectional imaging approaches play a key role in assessing bleeding brain injuries. Doctors commonly determine bleeding size and severity in CT and MRI. Separating and identifying artifacts is extremely important in processing medical images. Image and signal processing are used to classify tissues within images closely linked to edges. In CT images, a subjective process takes a stroke ‘s manual contour with less precision. This chapter presents the application of both image and signal processing techniques in the characterization of Brain Stroke field. This chapter also summarizes how to characterize the brain stroke using different image processing algorithms such as ROI based segmentation and watershed methods.

A deep learning based automatic segmentation approach for anatomical structures in intensity modulation radiotherapy

<abstract> <sec><title>Objective</title><p>To evaluate the automatic segmentation approach for organ at risk (OARs) and compare the parameters of dose volume histogram (DVH) in radiotherapy. Methodology: Thirty-three patients were selected to contour OARs using automatic segmentation approach which based on U-Net, applying them to a number of the nasopharyngeal carcinoma (NPC), breast, and rectal cancer respectively. The automatic contours were transferred to the Pinnacle System to evaluate contour accuracy and compare the DVH parameters.</p> </sec> <sec><title>Results</title><p>The time for manual contour was 56.5 ± 9, 23.12 ± 4.23 and 45.23 ± 2.39min for the OARs of NPC, breast and rectal cancer, and for automatic contour was 1.5 ± 0.23, 1.45 ± 0.78 and 1.8 ± 0.56 min. Automatic contours of Eye with the best Dice-similarity coefficients (DSC) of 0.907 ± 0.02 while with the poorest DSC of 0.459 ± 0.112 of Spinal Cord for NPC; And Lung with the best DSC of 0.944 ± 0.03 while with the poorest DSC of 0.709 ± 0.1 of Spinal Cord for breast; And Bladder with the best DSC of 0.91 ± 0.04 while with the poorest DSC of 0.43 ± 0.1 of Femoral heads for rectal cancer. The contours of Spinal Cord in H &amp; N had poor results due to the division of the medulla oblongata. The contours of Femoral head, which different from what we expect, also due to manual contour result in poor DSC.</p> </sec> <sec><title>Conclusion</title><p>The automatic contour approach based deep learning method with sufficient accuracy for research purposes. However, the value of DSC does not fully reflect the accuracy of dose distribution, but can cause dose changes due to the changes in the OARs volume and DSC from the data. Considering the significantly time-saving and good performance in partial OARs, the automatic contouring also plays a supervisory role.</p> </sec> </abstract>

Interim FDG-PET imaging during neoadjuvant chemoradiotherapy for esophageal cancer: Correlation with pathologic response.

175 Background: In this prospective study we evaluated whether changes in metabolic tumor parameters on interim flurodeoxyglucose positron emission tomography (FDG-PET) performed during neoadjuvant chemoradiotherapy (CRT) for esophageal cancer correlates with histopathologic tumor response. Methods: From February 2012 to February 2016, 60 patients with esophageal cancer underwent PET scans before therapy and after 30-36 Gy. Patients who underwent surgery after carboplatin/paclitaxel CRT were eligible for the current analysis. PET metrics of the primary site including maximum standardized uptake value (SUVmax), SUV mean, metabolic tumor volume (MTV), and total lesion glycolysis (TLG) were extracted from the pre-treatment and interim PET based on a manual contour and SUV 2.5 threshold. Patients were called histopathologic responders if they had a complete or near complete tumor response based on the modified Ryan scheme. Relative changes in PET metrics between pre-treatment and interim PET were compared between histopathologic responders and non-responders using the Mann-Whitney test and binary logistic regression. Results: Twenty-six patients were included in the analysis. Adenocarcinoma was the most common histology (n = 23). Eleven patients (42%) had a complete or near complete pathologic response to CRT (histopathologic responders). Changes in PET metrics from pre-treatment to interim PET based on the manual contour were not significantly different between responding and nonresponding tumors. The relative reduction of SUVmax (Mean ± SD) was 38.2% ± 28.4% for histopathologic responders and 27.9% ± 31.4% for non-responders. The relative reduction in MTV, SUV mean and TLG was 36.1% ± 26.2%, 23.5% ± 21.3%, and 49.3% ± 28.3% for histopathologic responders and 28.6% ± 32.0%, 11.8% ± 19.1%, and 33.1% ± 38.5% for histopathologic non-responders, respectively. When analyzed based on the SUV 2.5 threshold there continued to be no significant difference in PET metrics. Conclusions: In this pilot study we observed changes in metabolic tumor parameters on PET performed during CRT for esophageal cancer. However, these changes did not predict for histopathologic responders.

Accuracy of manual and automated rectal contours using helical tomotherapy image guidance scans during prostate radiotherapy.

94 Background: Prostate radiotherapy can be delivered using daily image-guided helical tomotherapy. Previous work has shown that contouring the rectum on the kV planning CT scan has a Jaccard conformity index (JCI) of 0.78 for different oncologists (inter-observer variability) and 0.82 for a single oncologist (intra-observer variability) (Lutgendorf-Caucig C et al. Feasibility of CBCT-based target and normal structure delineation in prostate cancer radiotherapy: multi-observer and image multi-modality study. Radiother Oncol. 2011;98(2):154-61.). Using the daily image guidance MV CT scan we have developed automated methods to contour the rectum in order to investigate the dose delivered over a course of treatment. We sought to quantify the accuracy of MV manual and automated contours. Methods: A single oncologist (JES) contoured the rectum on 370 MV scans for 10 participants treated with helical tomotherapy to prostate and pelvic lymph nodes. Accuracy of MV manual contours was tested using a scalar algorithm to enlarge and reduce the contours and intra-observer re-contouring at a 3-month interval. Automated contouring, incorporating the Chan-Vese algorithm, was developed and outputs were compared with manual contours. Results: JES could identify differences in MV manual contour size at the level of ±2.2 mm, equivalent to 1.7 pixels. The median JCI for MV re-contouring was 0.87 with inter-quartile range (IQR) 0.78 to 0.90. When compared with manual contours, automated outputs had a median JCI of 0.79 (IQR 0.74 to 0.79). These results were obtained after 3 iterations, each taking less than 10 seconds. Conclusions: Manual contouring using MV scans was accurate, at a level of approximately 2 mm, and reproducible, with JCI of 0.87. The time taken to contour was approximately 20 minutes per scan. Automated contouring was also reproducible with JCI of 0.79 and, in contrast, took less than a minute per scan. Both manual and automated methods produced results comparable to those for contouring using kV scans. We plan to use auto-contouring to calculate accumulated dose to the rectum in an initial cohort of 100 participants. These doses will be correlated with toxicity as part of the VoxTox Study.