Atlas-based Auto-segmentation Research Articles

Retrospective Comparison of Geometrical Accuracy among Atlas-based Auto-segmentation, Deep Learning Auto-segmentation, and Deformable Image Registration in the Treatment Replanning for Adaptive Radiotherapy of Head-and-Neck Cancer

Aims: This study aimed to evaluate the geometrical accuracy of atlas-based auto-segmentation (ABAS), deformable image registration (DIR), and deep learning auto-segmentation (DLAS) in adaptive radiotherapy (ART) for head-and-neck cancer (HNC). Subjects and Methods: Seventeen patients who underwent replanning for ART were retrospectively studied, and delineated contours on their replanning computed tomography (CT2) images were delineated. For DIR, the planning CT image (CT1) of the evaluated patients was utilized. In contrast, ABAS was performed using an atlas dataset comprising 30 patients who were not part of the evaluated group. DLAS was trained with 143 patients from different patients from the evaluated patients. The ABAS model was improved, and a modified ABAS (mABAS) was created by adding the evaluated patients’ own CT1 to the atlas datasets of ABAS (number of patients of the atlas dataset, 31). The geometrical accuracy of DIR, DLAS, ABAS, and mABAS was evaluated. Results: The Dice similarity coefficient in DIR was the highest, at >0.8 at all organs at risk. The mABAS was delineated slightly more accurately than the standard ABAS. There was no significant difference between ABAS and DLAS in delineation accuracy. DIR had the lowest Hausdorff distance (HD) value (within 10 mm). The HD values in ABAS, mABAS, and DLAS were within 16 mm. Conclusions: DIR delineation is the most geometrically accurate ART for HNC.

Enhancing the Contouring Efficiency for Head and Neck Cancer Radiotherapy Using Atlas-based Auto-segmentation and Scripting.

Intensity-modulated radiation therapy can deliver a highly conformal dose to a target while minimizing the dose to the organs at risk (OARs). Delineating the contours of OARs is time-consuming, and various automatic contouring software programs have been employed to reduce the delineation time. However, some software operations are manual, and further reduction in time is possible. This study aimed to automate running atlas-based auto-segmentation (ABAS) and software operations using a scripting function, thereby reducing work time. Dice coefficient and Hausdorff distance were used to determine geometric accuracy. The manual delineation, automatic delineation, and modification times were measured. While modifying the contours, the degree of subjective correction was rated on a four-point scale. The model exhibited generally good geometric accuracy. However, some OARs, such as the chiasm, optic nerve, retina, lens, and brain require improvement. The average contour delineation time was reduced from 57 to 29 min (p<0.05). The subjective revision degree results indicated that all OARs required minor modifications; only the submandibular gland, thyroid, and esophagus were rated as modified from scratch. The ABAS model and scripted automation in head and neck cancer reduced the work time and software operations. The time can be further reduced by improving contour accuracy.

Clinical evaluation of deep learning and atlas-based auto-segmentation for organs at risk delineation

Manual delineation of organs at risk and clinical target volumes is essential in radiotherapy planning. Atlas-based auto-segmentation (ABAS) algorithms have become available and been shown to provide accurate contouring for various anatomical sites. Recently, deep learning auto-segmentation (DL-AS) algorithms have emerged as the state-of-the-art in medical image segmentation. This study aimed to evaluate the effect of auto-segmentation on the clinical workflow for contouring different anatomical sites of cancer, such as head and neck (H&N), breast, abdominal region, and prostate. Patients with H&N, breast, abdominal, and prostate cancer (n = 30 each) were enrolled in the study. Twenty-seven different organs at four sites were evaluated. RayStation was used to apply the ABAS. Siemens AI-Rad Companion Organs RT was used to apply the DL-AS. Evaluations were performed with similarity indices using geometric methods, time-evaluation, and qualitative scoring visual evaluations by radiation oncologists. The DL-AS algorithm was more accurate than ABAS algorithm on geometric indices for half of the structures. The qualitative scoring results of the two algorithms were significantly different, and DL-AS was more accurate on many contours. DL-AS had 41%, 29%, 86%, and 15% shorter edit times in the HnN, breast, abdomen, and prostate groups, respectively, than ABAS. There were no correlations between the geometric indices and visual assessments. The time required to edit the contours was considerably shorter for DL-AS than for ABAS. Auto-segmentation with deep learning could be the first step for clinical workflow optimization in radiotherapy.

Clinical evaluation on automatic segmentation results of convolutional neural networks in rectal cancer radiotherapy.

Image segmentation can be time-consuming and lacks consistency between different oncologists, which is essential in conformal radiotherapy techniques. We aimed to evaluate automatic delineation results generated by convolutional neural networks (CNNs) from geometry and dosimetry perspectives and explore the reliability of these segmentation tools in rectal cancer. Forty-seven rectal cancer cases treated from February 2018 to April 2019 were randomly collected retrospectively in our cancer center. The oncologists delineated regions of interest (ROIs) on planning CT images as the ground truth, including clinical target volume (CTV), bladder, small intestine, and femoral heads. The corresponding automatic segmentation results were generated by DeepLabv3+ and ResUNet, and we also used Atlas-Based Autosegmentation (ABAS) software for comparison. The geometry evaluation was carried out using the volumetric Dice similarity coefficient (DSC) and surface DSC, and critical dose parameters were assessed based on replanning optimized by clinically approved or automatically generated CTVs and organs at risk (OARs), i.e., the Planref and Plantest. Pearson test was used to explore the correlation between geometric metrics and dose parameters. In geometric evaluation, DeepLabv3+ performed better in DCS metrics for the CTV (volumetric DSC, mean = 0.96, P< 0.01; surface DSC, mean = 0.78, P< 0.01) and small intestine (volumetric DSC, mean = 0.91, P< 0.01; surface DSC, mean = 0.62, P< 0.01), ResUNet had advantages in volumetric DSC of the bladder (mean = 0.97, P< 0.05). For critical dose parameters analysis between Planref and Plantest, there was a significant difference for target volumes (P< 0.01), and no significant difference was found for the ResUNet-generated small intestine (P > 0.05). For the correlation test, a negative correlation was found between DSC metrics (volumetric, surface DSC) and dosimetric parameters (δD95, δD95, HI, CI) for target volumes (P< 0.05), and no significant correlation was found for most tests of OARs (P > 0.05). CNNs show remarkable repeatability and time-saving in automatic segmentation, and their accuracy also has a certain potential in clinical practice. Meanwhile, clinical aspects, such as dose distribution, may need to be considered when comparing the performance of auto-segmentation methods.

Clinical evaluation of atlas-based auto-segmentation in breast and nodal radiotherapy.

Accurate contouring of anatomical structures allows for high-precision radiotherapy planning, targeting the dose at treatment volumes and avoiding organs at risk. Manual contouring is time-consuming with significant user variability, whereas auto-segmentation (AS) has proven efficiency benefits but requires editing before treatment planning. This study investigated whether atlas-based AS (ABAS) accuracy improves with template atlas group size and character-specific atlas and test case selection. One clinician retrospectively contoured the breast, nodes, lung, heart, and brachial plexus on 100 CT scans, adhering to peer-reviewed guidelines. Atlases were clustered in group sizes, treatment positions, chest wall separations, and ASs created with Mirada software. The similarity of ASs compared to reference contours was described by the Jaccard similarity coefficient (JSC) and centroid distance variance (CDV). Across group sizes, for all structures combined, the mean JSC was 0.6 (SD 0.3, p = .999). Across atlas-specific groups, 0.6 (SD 0.3, p = 1.000). The correlation between JSC and structure volume was weak in both scenarios (adjusted R2-0.007 and 0.185).Mean CDV was similar across groups but varied up to 1.2 cm for specific structures. Character-specific atlas groups and test case selection did not improve accuracy outcomes. High-quality ASs were obtained from groups containing as few as ten atlases, subsequently simplifying the application of ABAS. CDV measures indicating auto-segmentation variations on the x, y, and z axes can be utilised to decide on the clinical relevance of variations and reduce AS editing. High-quality ABASs can be obtained from as few as ten template atlases.Atlas and test case selection do not improve AS accuracy.Unlike well-known quantitative similarity indices, volume displacement metrics provide information on the location of segmentation variations, helping assessment of the clinical relevance of variations and reducing clinician editing. Volume displacement metrics combined with the qualitative measure of clinician assessment could reduce user variability.

Investigation of autosegmentation techniques on T2-weighted MRI for off-line dose reconstruction in MR-linac workflow for head and neck cancers.

In order to accurately accumulate delivered dose for head and neck cancer patients treated with the Adapt to Position workflow on the 1.5T magnetic resonance imaging (MRI)-linear accelerator (MR-linac), the low-resolution T2-weighted MRIs used for daily setup must be segmented to enable reconstruction of the delivered dose at each fraction. In this pilot study, we evaluate various autosegmentation methods for head and neck organs at risk (OARs) on on-board setup MRIs from the MR-linac for off-line reconstruction of delivered dose. Seven OARs (parotid glands, submandibular glands, mandible, spinal cord, and brainstem) were contoured on 43 images by seven observers each. Ground truth contours were generated using a simultaneous truth and performance level estimation (STAPLE) algorithm. Twenty total autosegmentation methods were evaluated in ADMIRE: 1-9) atlas-based autosegmentation using a population atlas library (PAL) of 5/10/15 patients with STAPLE, patch fusion (PF), random forest (RF) for label fusion; 10-19) autosegmentation using images from a patient's 1-4 prior fractions (individualized patient prior [IPP]) using STAPLE/PF/RF; 20) deep learning (DL) (3D ResUNet trained on 43 ground truth structure sets plus 45 contoured by one observer). Execution time was measured for each method. Autosegmented structures were compared to ground truth structures using the Dice similarity coefficient, mean surface distance (MSD), Hausdorff distance (HD), and Jaccard index (JI). For each metric and OAR, performance was compared to the inter-observer variability using Dunn's test with control. Methods were compared pairwise using the Steel-Dwass test for each metric pooled across all OARs. Further dosimetric analysis was performed on three high-performing autosegmentation methods (DL, IPP with RF and 4 fractions [IPP_RF_4], IPP with 1 fraction [IPP_1]), and one low-performing (PAL with STAPLE and 5 atlases [PAL_ST_5]). For five patients, delivered doses from clinical plans were recalculated on setup images with ground truth and autosegmented structure sets. Differences in maximum and mean dose to each structure between the ground truth and autosegmented structures were calculated and correlated with geometric metrics. DL and IPP methods performed best overall, all significantly outperforming inter-observer variability and with no significant difference between methods in pairwise comparison. PAL methods performed worst overall; most were not significantly different from the inter-observer variability or from each other. DL was the fastest method (33s per case) and PAL methods the slowest (3.7-13.8 min per case). Execution time increased with a number of prior fractions/atlases for IPP and PAL. For DL, IPP_1, and IPP_RF_4, the majority (95%) of dose differences were within±250cGy from ground truth, but outlier differences up to 785cGy occurred. Dose differences were much higher for PAL_ST_5, with outlier differences up to 1920cGy. Dose differences showed weak but significant correlations with all geometric metrics (R2 between 0.030 and 0.314). The autosegmentation methods offering the best combination of performance and execution time are DL and IPP_1. Dose reconstruction on on-board T2-weighted MRIs is feasible with autosegmented structures with minimal dosimetric variation from ground truth, but contours should be visually inspected prior to dose reconstruction in an end-to-end dose accumulation workflow.

Evaluation of Atlas-based auto-segmentation of liver in MR images for Yttrium-90 selective internal radiation therapy.

The aim was to explore the feasibility of applying an atlas-based auto-segmentation tool, MIM Atlas Segment, for liver delineation in MR images in Y-90 selective internal radiation therapy (SIRT). MR images of 41 liver patients treated with resin Y-90 SIRT were included: 20 patients' images were used to create an atlas, and the other 21 patients' images were used for testing. Auto-segmentation of liver in the MR images was performed with MIM Atlas Segment, and various settings for the auto-segmentation (i.e., with and without normalized deformable registration, single atlas-match and multi-atlas match, and multi-atlas match using different finalization methods) were tested. Auto-segmented liver contours were compared with physician manually-delineated contours, using Dice similarity coefficient (DSC) and mean distance to agreement (MDA). Ratio of volume (RV) and ratio of activity (RA) were calculated to further evaluate the auto-segmentation results. Auto-segmentations with normalized deformable registration generated better contours than those without normalized deformable registration. With normalized deformable registration, 3-atlas match using Majority Vote (MV) method generated better results than single-atlas match and 3-atlas match using STAPLE method, and generated similar results as 5-atlas match using MV method or STAPLE method. The average DSC, MDA, and RV of the contours generated with normalized deformable registration are 0.80-0.83, 0.60-0.67, and 0.91-1.00 cm, respectively. The average RA are 1.00-1.01, which indicate that the activities calculated using the auto-segmented liver contours are close to the accurate activities. The atlas-based auto-segmentation can be applied to generate initial liver contours in MR images for resin Y-90 SIRT, which can be used for activity calculations after physicians review.

Optimized Atlas-Based Auto-Segmentation of Bony Structures from Whole-Body Computed Tomography

To develop and test a method for fully automated segmentation of bony structures from whole-body CT and evaluate its performance compared to manual segmentation METHODS AND MATERIALS;: We developed a workflow for automatic whole-body bone segmentation using atlas-based segmentation method with a post-processing module (ABSPP) in MIM MAESTRO software. 52 CT scans comprised the training set to build the atlas library and 29 CT scans comprised the test set. To validate the workflow, we compared Dice similarity coefficient (DSC), mean distance to agreement (MDA), and relative volume errors (ΔV) between ABSPP and atlas-based segmentation with no post-processing (ABSNPP), with manual segmentation as the reference (gold standard). RESULTS;: The ABSPP method resulted in significantly improved segmentation accuracy (DSC range: 0.85 to 0.98) compared to the ABSNPP method (DSC range: 0.55 to 0.87) (p<0.001). MDA results also indicated high agreement between ABSPP and manual reference delineations (range: 0.11-1.56 mm), which was significantly improved compared to ABSNPP (range: 1.00-2.34 mm) for for the majority of tested bony structures. Relative volume errors were also significantly lower for ABSPP compared to ABSNPP for most bony structures. We developed a fully automated MIM workflow for bony structure segmentation from whole-body CT, which exhibited high accuracy compared to manual delineation. The integrated post-processing module significantly improved workflow performance.

Continuing professional development - Radiation Therapy.

Journal of Medical Radiation SciencesEarly View Continuing Professional DevelopmentOpen Access Continuing professional development - Radiation Therapy First published: 08 February 2023 https://doi.org/10.1002/jmrs.655AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Maximise your CPD by reading the following selected article and answer the five questions. Please remember to self-claim your CPD and retain your supporting evidence. Answers will be available via the QR code and online at www.asmirt.org/news-and-publications/jmrs, as well as published in JMRS – Volume 70, Issue 4 December 2023. Radiation Therapy – Original Article Clinical evaluation of deep learning and atlas-based auto-segmentation for critical organs at risk in radiation therapy Gibbons E, Hoffmann M, Westhuyzen J, Hodgson A, Chick B, Last A. (2023) J Med Radiat Sci. https://doi.org/10.1002/jmrs.618 How are deep learning (DL) auto-segmentation contours produced on an incoming data set? DL models use deformable image registration to transform contours from a pre-defined library of similar data sets DL models predict contour location using optimised convolution ‘neural’ networks (CNNs) that have been trained to identify complex non-linear spatial relationships within data sets DL models generate contours by estimating spatial position on a data set using the Dice similarity coefficient (DSC) formula DL models require users to roughly contour the incoming data set by hand before the algorithm can fully optimise each organ at risk (OAR) Which of the following is correct in terms of the number of training data sets used for atlas and DL segmentation? Atlas-based auto-segmentation typically uses a larger number of training data sets than DL models If an atlas and DL auto-segmentation model are trained using the same number of data sets, they are likely to output identical contours The larger number of training data sets used in DL segmentation naturally incorporates a more diverse range of patient anatomy than atlas-based methods DL auto-segmentation models typically reach a performance plateau when trained on 10–30 data sets Which DL-based OARs experienced statistically significant time improvements over atlas-based OARs? Spinal Cord, Rectum, Bladder, Left Femoral Head Spinal Cord, Oral Cavity, Left Lung, Oesophagus, Heart Left Parotid, Spinal Cord, Left Lung, Rectum, Bladder, Left Femoral Head Left Parotid, Rectum, Bladder, Left Femoral Head For auto-segmented OAR contours, it is recommended that: All DL generated OARs should be reviewed by an expert clinician prior to clinical use All DL generated OARs are accurate enough to be used clinically without human review Only a specific list of DL generated OARs requires expert review prior to clinical use Only a specific list of atlas generated OARs require expert review prior to clinical use Why can vendor trained ‘generic’ DL models be beneficial to a radiation therapy workflow? They will always output contours that are quantitatively more accurate than a DL model trained by an individual department They are pre-optimised to output contours that will meet the specific contouring protocols for all departments worldwide The output contours will always require less editing time than a DL model trained by an individual department They can output quality contours without the time, data or resource requirements needed for departments to individually train their own DL models Recommended further reading: 1Brouwer CL, Boukerroui D, Oliveira J, et al. Assessment of manual adjustment performed in clinical practice following deep learning contouring for head and neck organs at risk in radiotherapy. Phys Imaging Radiat Oncol. 2020; 16: 54– 60. 2Sherer MV, Lin D, Elguindi S, et al. Metrics to evaluate the performance of auto-segmentation for radiation treatment planning: A critical review. Radiother Oncol. 2021; 160: 185– 91. 3van Dijk LV, Van den Bosch L, Aljabar P, et al. Improving automatic delineation for head and neck organs at risk by Deep Learning Contouring. Radiother Oncol. 2020; 142: 115– 23. Answers Scan this QR code to find the answers, or visit www.asmirt.org/news-and-publications/jmrs Early ViewOnline Version of Record before inclusion in an issue ReferencesRelatedInformation

Clinical Feasibility of Commercial Artificial Intelligence-Based Auto Contouring of Target Volumes and Organs-at-Risk in Breast Cancer Patients

<h3>Purpose/Objective(s)</h3> Manual clinical target volume (CTV) and organ at-risk (OAR) segmentation for planning remains a significant component of physician workload. In addition, inter-physician segmentation variability can introduce uncertainty in planning process. While current atlas-based auto-segmentation has proven to be useful for contouring, it typically requires significant upfront time investment and is physician dependent. Recently, artificial-intelligence-based contouring have been made commercially available. These systems are physician-independent, and rely on machine learning rather than a user-selected atlas of patient datasets. In this study, we evaluated the potential benefits of a commercially available artificial intelligence-based auto contouring (AI-AC) system by comparing AI-AC generated and manually delineated contours for breast patients. <h3>Materials/Methods</h3> CTVs (ipsilateral breast) and OARs (contralateral breast, heart, lungs, esophagus, and spinal cord) were manually contouring (MC) on planning computed tomography scans of 10 breast cancer patients (5 patients received breast-conserving surgery). Commercial AI auto-contouring software (AI-Rad Companion Organs RT) was then used to generate the same target and risk volumes. Volumes were compared using quantitative metrics, including the Dice Similarity Coefficient (DSC), and mean distance agreement (MDA). <h3>Results</h3> Correlation between the AI-AC and MC generated volumes was strong for the evaluated OARs, with a mean DSC higher than 0.90 for all OARs except the esophagus (mean DSC 0.83 ± 0.09)). In addition, Comparison of CTV volumes also showed excellent results, with a mean DSC higher than 0.93. For the OARs, mean DSC values ranged from 0.98 ± 0.01 to 0.83 ± 0.09, while the mean MDA ranged from 0.68 ± 0.3 to 0.84 +0.54 mm to mm. [Similar results were achieved for the CTV, with the mean DSC ranging from 0.93 ± 0.03 to 0.94 ± 0.06 to 0.93, and mean MDA from 1.71 ± 0.91 mm to 1.76 ± 0.68 mm. AI-AC generated volumes were produced in approximately 2 min, while MC of the same volumes takes the physician an estimated 30 min. <h3>Conclusion</h3> The feasibility of commercial AI-AC in breast RT planning was demonstrated. Immediately on activation, the AI-AC system was able to produce volumes showing significant agreement with physician-generated volumes. While AI-AC-generated volumes may require minor edits by the physician, clinical efficiency is substantially improved. Furthermore, initial volumes created by AI-AC are physician-independent, and may help to improve contour consistency in clinic compared to strictly manual approaches.

Clinical evaluation of deep learning and atlas-based auto-segmentation for critical organs at risk in radiation therapy.

Contouring organs at risk (OARs) is a time-intensive task that is a critical part of radiation therapy. Atlas-based automatic segmentation has shown some success at reducing this time burden on practitioners; however, this method often requires significant manual editing to reach a clinically accurate standard. Deep learning (DL) auto-segmentation has recently emerged as a promising solution. This study compares the accuracy of DL and atlas-based auto-segmentation in relation to clinical 'gold standard' reference contours. Ninety CT datasets (30 head and neck, 30 thoracic, 30 pelvic) were automatically contoured using both atlas and DL segmentation techniques. Sixteen critical OARs were then quantitatively measured for accuracy using the Dice similarity coefficient (DSC) and Hausdorff distance (HD). Qualitative analysis was performed to visually classify the accuracy of each structure into one of four explicitly defined categories. Additionally, the time to edit atlas and DL contours to a clinically acceptable level was recorded for a subset of 9 OARs. Of the 16 OARs analysed, DL delivered statistically significant improvements over atlas segmentation in 13 OARs measured with DSC, 12 OARs measured with HD, and 12 OARs measured qualitatively. The mean editing time for the subset of DL contours was 50%, 23% and 61% faster (all P < 0.05) than that of atlas segmentation for the head and neck, thorax, and pelvis respectively. Deep learning segmentation comprehensively outperformed atlas-based contouring for the majority of evaluated OARs. Improvements were observed in geometric accuracy and visual acceptability, while editing time was reduced leading to increased workflow efficiency.

The clinical evaluation of atlas-based auto-segmentation for automatic contouring during cervical cancer radiotherapy.

PurposeOur purpose was to investigate the influence of atlas library size and CT cross-slice number on the accuracy and efficiency of the atlas-based auto-segmentation (ABAS) method for the automatic contouring of clinical treatment volume (CTV) and organs at risk (OARs) during cervical cancer radiotherapy.MethodsOf 140 cervical cancer patients, contours from 20, 40, 60, 80, 100, and 120 patients were selected incrementally to create six atlas library groups in ABAS. Another 20 tested patients were automatically contoured with the ABAS method and manually contoured by the same professional oncologist. Contours included CTV, bladder, rectum, femoral head-L, femoral head-R, and spinal cord. The CT cross-slice numbers of the 20 tested patients included 61, 65, 72, 75, 81, and 84. The index of dice similarity coefficients (DSCs) and Hausdorff distance (HD) were used to assess the consistency between ABAS automatic contouring and manual contouring. The randomized block analysis of variance and paired t-test were used for statistical analysis.ResultsThe mean DSC values of “CTV, bladder, femoral head, and spinal cord” were all larger than 0.8. The femoral head and spinal cord showed a high degree of agreement between ABAS automatic contouring and manual contouring, with a mean DC >0.80 and HD <1 cm in all atlas library groups. A post-hoc least significant difference comparison indicated that no significant difference had been found between different atlas library sizes with DSC and HD values. For ABAS efficiency, the atlas library size had no effect on the time of ABAS automatic contouring. The time of automatic contouring increased slightly with the increase in CT cross-slice numbers, which were 99.9, 106.8, 114.0, 120.6, 127.9, and 134.8 s with CT cross-slices of 61, 65, 72, 75, 81, and 84, respectively.ConclusionA total of 20 atlas library sizes and a minimum CT cross-slice number including CTV and OARs are enough for ensuring the accuracy and efficiency of ABAS automatic contouring during cervical cancer radiotherapy.

PD-0498 Atlas-based auto-segmentation of pelvic CTVs in FDG-PET-CT planning of anal cancer

PD-0498 Atlas-based auto-segmentation of pelvic CTVs in FDG-PET-CT planning of anal cancer

Evaluation Exploration of Atlas-Based and Deep Learning-Based Automatic Contouring for Nasopharyngeal Carcinoma.

PurposeThe purpose of this study was to evaluate and explore the difference between an atlas-based and deep learning (DL)-based auto-segmentation scheme for organs at risk (OARs) of nasopharyngeal carcinoma cases to provide valuable help for clinical practice.Methods120 nasopharyngeal carcinoma cases were established in the MIM Maestro (atlas) database and trained by a DL-based model (AccuContour®), and another 20 nasopharyngeal carcinoma cases were randomly selected outside the atlas database. The experienced physicians contoured 14 OARs from 20 patients based on the published consensus guidelines, and these were defined as the reference volumes (Vref). Meanwhile, these OARs were auto-contoured using an atlas-based model, a pre-built DL-based model, and an on-site trained DL-based model. These volumes were named Vatlas, VDL-pre-built, and VDL-trained, respectively. The similarities between Vatlas, VDL-pre-built, VDL-trained, and Vref were assessed using the Dice similarity coefficient (DSC), Jaccard coefficient (JAC), maximum Hausdorff distance (HDmax), and deviation of centroid (DC) methods. A one-way ANOVA test was carried out to show the differences (between each two of them).ResultsThe results of the three methods were almost similar for the brainstem and eyes. For inner ears and temporomandibular joints, the results of the pre-built DL-based model are the worst, as well as the results of atlas-based auto-segmentation for the lens. For the segmentation of optic nerves, the trained DL-based model shows the best performance (p < 0.05). For the contouring of the oral cavity, the DSC value of VDL-pre-built is the smallest, and VDL-trained is the most significant (p < 0.05). For the parotid glands, the DSC of Vatlas is the minimum (about 0.80 or so), and VDL-pre-built and VDL-trained are slightly larger (about 0.82 or so). In addition to the oral cavity, parotid glands, and the brainstem, the maximum Hausdorff distances of the other organs are below 0.5 cm using the trained DL-based segmentation model. The trained DL-based segmentation method behaves well in the contouring of all the organs that the maximum average deviation of the centroid is no more than 0.3 cm.ConclusionThe trained DL-based segmentation performs significantly better than atlas-based segmentation for nasopharyngeal carcinoma, especially for the OARs with small volumes. Although some delineation results still need further modification, auto-segmentation methods improve the work efficiency and provide a level of help for clinical work.

Quantitative Comparisons of Deep-learning-based and Atlas-based Auto- segmentation of the Intermediate Risk Clinical Target Volume for Nasopharyngeal Carcinoma.

Manual segment target volumes were time-consuming and inter-observer variability couldn't be avoided. With the development of computer science, auto-segmentation had the potential to solve this problem. To evaluate the accuracy and stability of Atlas-based and deep-learning-based auto-segmentation of the intermediate risk clinical target volume, composed of CTV2 and CTVnd, for nasopharyngeal carcinoma quantitatively. A cascade-deep-residual neural network was constructed to automatically segment CTV2 and CTVnd by deep learning method. Meanwhile, a commercially available software was used to automatically segment the same regions by Atlas-based method. The datasets included contrast computed tomography scans from 102 patients. For each patient, the two regions were manually delineated by one experienced physician. The similarity between the two auto-segmentation methods was quantitatively evaluated by Dice similarity coefficient, the 95th Hausdorff distance, volume overlap error and relative volume difference, respectively. Statistical analyses were performed using the ranked Wilcoxon test. The average Dice similarity coefficient (±standard deviation) given by the deep-learning- based and Atlas-based auto-segmentation were 0.84 (±0.03) and 0.74 (±0.04) for CTV2, 0.79 (±0.02) and 0.68 (±0.03) for CTVnd, respectively. For the 95th Hausdorff distance, the corresponding values were 6.30±3.55 mm and 9.34±3.39 mm for CTV2, 7.09±2.27 mm and 14.33±3.98 mm for CTVnd. Besides, volume overlap error and relative volume difference could also predict the same situations. Statistical analyses showed significant difference between the two auto-segmentation methods (p<0.01). Compared with the Atlas-based segmentation approach, the deep-learning-based segmentation method performed better both in accuracy and stability for meaningful anatomical areas other than organs at risk.

Development of Simplified Auto-Segmentable Functional Cardiac Atlas

PurposeThere is increasing evidence that radiation doses to cardiac substructures are associated with cardiac adverse events. Manual delineation of cardiac substructures is time-consuming, and auto-segmentation of cardiac substructure atlases has consequently been evaluated. However, proper automatic delineation of small substructures, such as the left anterior descending coronary artery, is challenging, and auto-segmentation of cardiac conduction system substructures has never been evaluated, despite multiple reports of radiation-induced arrhythmia after thoracic irradiations. The aim of this study was to propose and evaluate a simplified auto-segmentable functional cardiac atlas. Methods and MaterialsWe created a cardiac substructure atlas based on 20 computed tomography scans from patients with breast cancer comprising the 4 cardiac cavities, a high-risk cardiac zone as a left anterior descending coronary artery surrogate, and the 2 cardiac conduction nodes. Automatic delineation of this atlas by an atlas-based auto-segmentation algorithm was evaluated on a validation data set, consisting of 20 additional computed tomography scans. Dice similarity coefficients were used to evaluate the concordance level between the manual and the automatic contours; a dosimetric comparison between mean and maximum doses to the manual and to the auto-segmented substructures was additionally performed, based on intensity modulated radiation therapy treatment plans of the patients of the validation set. ResultsAverage dice similarity coefficient values were 0.78 for the 4 cardiac cavities, 0.65 for the high-risk cardiac zones, 0.56 for the sinoatrial node, and 0.15 for the atrioventricular node. Compared with manual contours, auto-segmented substructures were slightly smaller but the dosimetric parameters were similar. ConclusionsWe proposed a simplified functional cardiac atlas that included the cardiac conduction system and circumvented coronary delineation difficulties by using a surrogate high-risk cardiac zone. Most cardiac substructures were associated with acceptable atlas-based auto-segmentation properties. Such an atlas could be used for epidemiologic studies and for clinical practice.

Abstract P3-19-14: Development of simplified auto-segmentable functional cardiac atlas

Abstract Objective: There are increasing evidences that radiation doses to specific cardiac substructures are associated with cardiac adverse events. Manual delineation of cardiac substructures is time consuming, and auto-segmentation of cardiac substructure atlases has consequently been evaluated. However, proper delineation of small substructures, such as the left anterior descending coronary artery (LADCA), is challenging and conduction system substructures have never been considered, despite frequent reports of radiation-induced arrhythmias for thoracic irradiation. The aim of this study was to propose and evaluate a simplified auto-segmentable functional cardiac atlas. Methods: We created a cardiac substructure atlas from 20 breast cancer patients’ CT scans consisting of the four cardiac cavities, of a high-risk cardiac zone (HRCZ) as a LADCA surrogate and of the two cardiac conduction nodes. Performance evaluation of atlas-based auto-segmentation (ABAS) was evaluated on a validation data set consisting of 20 additional CT scans. Dice similarity coefficients (DSC) were used to evaluate the concordance level between the manual and the automatic segmentations. Results: The average duration of manual segmentation of the proposed cardiac atlas ranged between fifteen to twenty minutes, while ABAS lasted two minutes. The median DSC for the delineated cardiac substructures was 0.718. The highest similarity between manual and automatic segmentation was observed for the left ventricle with a median DSC of 0.87, while the lowest similarity between manual and automatic segmentation was observed for the NAV with a median DSC of 0.15. Regardless of the considered cardiac substructure, auto-segmentation tended to result into smaller volumes than manual segmentation. While the volumes were not statistically different for the left ventricle, the volumes were statistically different for the other substructures. While smaller, the auto-segmented NAV and SAN were systematically localized within the manual contours. The auto-segmented NAV could be approximated by a 1.6-cm sphere and the auto-segmented SAN by a 1.0-cm sphere. Conclusion: We proposed a simplified functional cardiac atlas, for which coronary delineation difficulties were circumvent using a surrogate high risk cardiac zone and cardiac conduction system was considered. Most cardiac substructures were associated with acceptable ABAS properties. Such atlas could be potentially evaluated for epidemiological studies and clinical practice. Table 1.performance of atlas-based auto-segmentation of the proposed simplified cardiac atlasCardiac substructureDSC Overlap index [range]Cardiac cavitiesLeft ventricle0.87 [0.57-0.92]Right ventricle0.76 [0.45-0.85]Left atria0.77 [0.61-0.84]Right atria0.76 [0.54-0.85]High risk cardiac zone0.65 [0.38-0.79]Cardiac conduction systemSinoatrial node0.56 [0.27-0.67]Atrio-ventricular node0.15 [0.00-0.35] Citation Format: Pierre Loap, Ludovic De Marzi, Krassen Kirov, Vincent Servois, Alain Fourquet, Abdelhafidh Khoubeyb, Youlia Kirova. Development of simplified auto-segmentable functional cardiac atlas [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P3-19-14.

Evaluating Automatic Segmentation for Swallowing-Related Organs for Head and Neck Cancer.

Purpose: To evaluate the accuracy of deep-learning-based auto-segmentation of the superior constrictor, middle constrictor, inferior constrictor, and larynx in comparison with a traditional multi-atlas-based method. Methods and Materials: One hundred and five computed tomography image datasets from 83 head and neck cancer patients were retrospectively collected and the superior constrictor, middle constrictor, inferior constrictor, and larynx were analyzed for deep-learning versus multi-atlas-based segmentation. Eighty-three computed tomography images (40 diagnostic computed tomography and 43 planning computed tomography) were used for training the convolutional neural network, and for atlas-based model training. The remaining 22 computed tomography datasets were used for validation of the atlas-based auto-segmentation versus deep-learning-based auto-segmentation contours, both of which were compared with the corresponding manual contours. Quantitative measures included Dice similarity coefficient, recall, precision, Hausdorff distance, 95th percentile of Hausdorff distance, and mean surface distance. Dosimetric differences between the auto-generated contours and manual contours were evaluated. Subjective evaluation was obtained from 3 clinical observers to blindly score the autosegmented structures based on the percentage of slices that require manual modification. Results: The deep-learning-based auto-segmentation versus atlas-based auto-segmentation results were compared for the superior constrictor, middle constrictor, inferior constrictor, and larynx. The mean Dice similarity coefficient values for the 4 structures were 0.67, 0.60, 0.65, and 0.84 for deep-learning-based auto-segmentation, whereas atlas-based auto-segmentation has Dice similarity coefficient results at 0.45, 0.36, 0.50, and 0.70, respectively. The mean 95th percentile of Hausdorff distance (cm) for the 4 structures were 0.41, 0.57, 0.59, and 0.54 for deep-learning-based auto-segmentation, but 0.78, 0.95, 0.96, and 1.23 for atlas-based auto-segmentation results, respectively. Similar mean dose differences were obtained from the 2 sets of autosegmented contours compared to manual contours. The dose–volume discrepancies and the average modification rates were higher with the atlas-based auto-segmentation contours. Conclusion: Swallowing-related structures are more accurately generated with DL-based versus atlas-based segmentation when compared with manual contours.

Evaluation of an Atlas-Based Auto-Segmentation Tool of Target Volumes and Organs at Risk in Head and Neck Radiation Therapy

Evaluation of an Atlas-Based Auto-Segmentation Tool of Target Volumes and Organs at Risk in Head and Neck Radiation Therapy

Atlas-based auto-segmentation for delineating the heart and cardiac substructures in breast cancer radiation therapy

Background This study aimed to develop and validate an automatic multi-atlas segmentation method for delineating the heart and substructures in breast cancer radiation therapy (RT). Material and methods The atlas database consisted of non-contrast-enhanced planning CT scans from 42 breast cancer patients, each with one manual delineation of the heart and 22 cardiac substructures. Half of the patients were scanned during free-breathing, the rest were scanned during a deep inspiration breath-hold. The auto-segmentation was developed in the MIM software system and validated geometrically and dosimetrically in two steps: The first validation in a small dataset to ensure consistency of the atlas. This was succeeded by a final test where multiple manual delineations in CT scans of 12 breast cancer patients were compared to the auto-segmentation. For geometric evaluation, the dice similarity coefficient (DSC) and the mean surface distance (MSD) were used. For dosimetric evaluation, the RT doses to each substructure in the manual and the automatic delineations were compared. Results In the first validation, a high geometric and dosimetric performance between the automatic and manual delineations was observed for all substructures. The final test confirmed a high agreement between the automatic and manual delineations for the heart (DSC = 0.94) and the cardiac chambers (DSC: 0.75–0.86). The difference in MSD between the automatic and manual delineations was low (<4 mm) in all structures. Finally, a high correlation between mean RT doses for the automatic and the manual delineations was observed for the heart and substructures. Conclusions An automatic segmentation tool for delineation of the heart and substructures in breast cancer RT was developed and validated with a high correlation between the automatic and manual delineations. The atlas is pivotal for large-scale evaluations of radiation-associated heart disease.