95th Percentile Hausdorff Distance Research Articles

A deep learning-based auto-segmentation system for organs-at-risk on whole-body computed tomography images for radiation therapy

Background and purposeDelineating organs at risk (OARs) on computed tomography (CT) images is an essential step in radiation therapy; however, it is notoriously time-consuming and prone to inter-observer variation. Herein, we report a deep learning-based automatic segmentation (AS) algorithm (WBNet) that can accurately and efficiently delineate all major OARs in the entire body directly on CT scans. Materials and methodsWe collected 755 CT scans of the head and neck, thorax, abdomen, and pelvis and manually delineated 50 OARs on the CT images. The CT images with contours were split into training and test sets consisting of 505 and 250 cases, respectively, to develop and validate WBNet. The volumetric Dice similarity coefficient (DSC) and 95th-percentile Hausdorff distance (95% HD) were calculated to evaluate delineation quality for each OAR. We compared the performance of WBNet with three AS algorithms: one commercial multi-atlas-based automatic segmentation (ABAS) software, and two deep learning-based AS algorithms, namely, AnatomyNet and nnU-Net. We have also evaluated the time saving and dose accuracy of WBNet. ResultsWBNet achieved average DSCs of 0.84 and 0.81 on in-house and public datasets, respectively, which outperformed ABAS, AnatomyNet, and nnU-Net. WBNet could reduce the delineation time significantly and perform well in treatment planning, with clinically acceptable dose differences compared with those in manual delineation. ConclusionThis study shows the feasibility and benefits of using WBNet in clinical practice.

Evaluation of delineating the target volume by radiation therapists in breast cancer patients

AimBreast radiotherapy accounts for a significant workload in radiotherapy departments. In 2015 it became clinical practice at the Netherlands Cancer Institute for radiation therapists (RTTs) to delineate the clinical target volume of the breast tissue (CTVbreast) and in 2017 axilla level I-II (CTVln12) according to a delineation atlas. All RTTs were trained and got individual feedback. The aim of this retrospective study was to investigate the variation between the CTVbreast with or without CTVln12 delineated by a trained group of radiation therapists and the clinical adjusted delineations by the radiation oncologist/physician assistant (RO/PA), in a large group of patients treated between January 2017 and June 2020.Materials and Methods1012 computer tomography based delineations of CTVbreast and 146 of CTVln12 were collected from January 2017-June 2020. The RTT and RO/PA delineations were compared using the Dice coefficient and the 95th percentile Hausdorff Distance (95%HD). Statistical significance of the differences was tested using a Mann-Whitney test.ResultsDifferences in CTVbreast delineations were small. A median Dice score of 1.00 for all years, where 83% of the patients had a Dice score > 0.99. For CTVln12 the magnitude of edits made by RO/PAs decreased over time, with the Dice increasing from a median of 0.87 in 2017 to 0.90 in 2020 (p = 0.031). The 95%HD decreased from a median of 0.93 cm in 2017 to 0.61 cm in 2020 (p = 0.051).ConclusionsThis retrospective study shows that trained dedicated RTTs are capable in delivering the same quality delineations as RO/PAs. The low variability supports the increasing role of RTTs in the contouring process, likely making it more time efficient.

Automatic Segmentation of Clinical Target Volumes for Post-Modified Radical Mastectomy Radiotherapy Using Convolutional Neural Networks.

BackgroundThis study aims to construct and validate a model based on convolutional neural networks (CNNs), which can fulfil the automatic segmentation of clinical target volumes (CTVs) of breast cancer for radiotherapy.MethodsIn this work, computed tomography (CT) scans of 110 patients who underwent modified radical mastectomies were collected. The CTV contours were confirmed by two experienced oncologists. A novel CNN was constructed to automatically delineate the CTV. Quantitative evaluation metrics were calculated, and a clinical evaluation was conducted to evaluate the performance of our model.ResultsThe mean Dice similarity coefficient (DSC) of the proposed model was 0.90, and the 95th percentile Hausdorff distance (95HD) was 5.65 mm. The evaluation results of the two clinicians showed that 99.3% of the chest wall CTV slices could be accepted by clinician A, and this number was 98.9% for clinician B. In addition, 9/10 of patients had all slices accepted by clinician A, while 7/10 could be accepted by clinician B. The score differences between the AI (artificial intelligence) group and the GT (ground truth) group showed no statistically significant difference for either clinician. However, the score differences in the AI group were significantly different between the two clinicians. The Kappa consistency index was 0.259. It took 3.45 s to delineate the chest wall CTV using the model.ConclusionOur model could automatically generate the CTVs for breast cancer. AI-generated structures of the proposed model showed a trend that was comparable, or was even better, than those of human-generated structures. Additional multicentre evaluations should be performed for adequate validation before the model can be completely applied in clinical practice.

Validation of a CT-based motion model with in-situ fluoroscopy for lung surface deformation estimation

Many surrogate-based motion models (SMMs), proposed to guide motion management in radiotherapy, are constructed by correlating motion of an external surrogate and internal anatomy during CT-simulation. Changes in this correlation define model break down. We validate a methodology that incorporates fluoroscopic (FL) images acquired during treatment for SMM construction and update. Under a prospective IRB, 4DCT scans, VisionRT (VRT) surfaces, and orthogonal FLs were collected from five lung cancer patients. VRT surfaces and two FL time-series were acquired pre- and post-treatment. A simulated annealing optimization scheme was used to estimate optimal lung deformations by maximizing the mutual information (MI) between digitally reconstructed radiographs (DRRs) of the SMM-estimated 3D images and FLs. Our SMM used partial-least-regression and was trained using the optimal deformations and VRT surfaces from the first breathing-cycle. SMM performance was evaluated using the MI score between reference FLs and the corresponding SMM or phase-assigned 4DCT DRRs. The Hausdorff distance for contoured landmarks was used to evaluate target position estimation error. For four out of five patients, two principal components approximated lung surface deformations with submillimeter accuracy. Analysis of the MI score between more than 4000 pairs of FL and DRR demonstrated that our model led to more similarity between the FL and DRR images compared to 4DCT and DRR images from a model based on an a priori correlation model. Our SMM consistently displayed lower mean and 95th percentile Hausdorff distances. For one patient, 95th percentile Hausdorff distance was reduced by 11 mm. Patient-averaged reductions in mean and 95th percentile Hausdorff distances were 3.6 mm and 7 mm for right-lung, and 3.1 mm and 4 mm for left-lung targets. FL data were used to evaluate model performance and investigate the feasibility of model update. Despite variability in breathing, use of post-treatment FL preserved model fidelity and consistently outperformed 4DCT for position estimation.

Deep learning applications in automatic segmentation and reconstruction in CT-based cervix brachytherapy.

PurposeMotivated by recent advances in deep learning, the purpose of this study was to investigate a deep learning method in automatic segment and reconstruct applicators in computed tomography (CT) images for cervix brachytherapy treatment planning.Material and methodsU-Net model was developed for applicator segmentation in CT images. Sixty cervical cancer patients with Fletcher applicator were divided into training data and validation data according to ratio of 50 : 10, and another 10 patients with Fletcher applicator were employed to test the model. Dice similarity coefficient (DSC) and 95th percentile Hausdorff distance (HD95) were used to evaluate the model. Segmented applicator coordinates were calculated and applied into RT structure file. Tip error and shaft error of applicators were evaluated. Dosimetric differences between manual reconstruction and deep learning-based reconstruction were compared.ResultsThe averaged overall 10 test patients’ DSC, HD95, and reconstruction time were 0.89, 1.66 mm, and 17.12 s, respectively. The average tip error was 0.80 mm, and the average shaft error was less than 0.50 mm. The dosimetric differences between manual reconstruction and automatic reconstruction were 0.29% for high-risk clinical target volume (HR-CTV) D90%, and less than 2.64% for organs at risk D2cc at a scenario of doubled maximum shaft error.ConclusionsWe proposed a deep learning-based reconstruction method to localize Fletcher applicator in three-dimensional CT images. The achieved accuracy and efficiency confirmed our method as clinically attractive. It paves the way for the automation of brachytherapy treatment planning.

Development and validation of a deep learning algorithm for auto-delineation of clinical target volume and organs at risk in cervical cancer radiotherapy

PurposeThe delineation of the clinical target volume (CTV) is a crucial, laborious and subjective step in cervical cancer radiotherapy. The aim of this study was to propose and evaluate a novel end-to-end convolutional neural network (CNN) for fully automatic and accurate CTV in cervical cancer. MethodsA total of 237 computed tomography (CT) scans of patients with locally advanced cervical cancer were collected and evaluated. A novel 2.5D CNN network, called DpnUNet, was developed for CTV delineation and further applied for CTV and organ-at-risk (OAR) delineation simultaneously. Comprehensive comparisons and experiments were performed. The mean Dice similarity coefficient (DSC), 95th percentile Hausdorff distance (95HD) and subjective evaluation were used to assess the performance of this method. ResultsThe mean DSC and 95HD values were 0.86 and 5.34 mm for the delineated CTVs. The clinical experts’ subjective assessments showed that 90% of the predicted contours were acceptable for clinical usage. The mean DSC and 95HD values were 0.91 and 4.05 mm for bladder, 0.85 and 2.16 mm for bone marrow, 0.90 and 1.27 mm for left femoral head, 0.90 and 1.51 mm for right femoral head, 0.82 and 4.29 mm for rectum, 0.85 and 4.35 mm for bowel bag, 0.82 and 4.96 mm for spinal cord respectively. The average delineation time for one patient’s CT images was within 15 seconds. ConclusionThe experimental results demonstrate that the CTV and OARs delineated for cervical cancer by DpnUNet was in close agreement with the ground truth. DpnUNet could significantly reduce the radiation oncologists’ contouring time.

Gross tumor volume segmentation for head and neck cancer radiotherapy using deep dense multi-modality network

In radiation therapy, the accurate delineation of gross tumor volume (GTV) is crucial for treatment planning. However, it is challenging for head and neck cancer (HNC) due to the morphology complexity of various organs in the head, low targets to background contrast and potential artifacts on conventional planning CT images. Thus, manual delineation of GTV on anatomical images is extremely time consuming and suffers from inter-observer variability that leads to planning uncertainty. With the wide use of PET/CT imaging in oncology, complementary functional and anatomical information can be utilized for tumor contouring and bring a significant advantage for radiation therapy planning. In this study, by taking advantage of multi-modality PET and CT images, we propose an automatic GTV segmentation framework based on deep learning for HNC. The backbone of this segmentation framework is based on 3D convolution with dense connections which enables a better information propagation and takes full advantage of the features extracted from multi-modality input images. We evaluate our proposed framework on a dataset including 250 HNC patients. Each patient receives both planning CT and PET/CT imaging before radiation therapy (RT). Manually delineated GTV contours by radiation oncologists are used as ground truth in this study. To further investigate the advantage of our proposed Dense-Net framework, we also compared with the framework using 3D U-Net which is the state-of-the-art in segmentation tasks. Meanwhile, for each frame, the performance comparison between single modality input (PET or CT image) and multi-modality input (both PET/CT) is conducted. Dice coefficient, mean surface distance (MSD), 95th-percentile Hausdorff distance (HD95) and displacement of mass centroid (DMC) are calculated for quantitative evaluation. The dataset is split into train (140 patients), validation (35 patients) and test (75 patients) groups to optimize the network. Based on the results on independent test group, our proposed multi-modality Dense-Net (Dice 0.73) shows better performance than the compared network (Dice 0.71). Furthermore, the proposed Dense-Net structure has less trainable parameters than the 3D U-Net, which reduces the prediction variability. In conclusion, our proposed multi-modality Dense-Net can enable satisfied GTV segmentation for HNC using multi-modality images and yield superior performance than conventional methods. Our proposed method provides an automatic, fast and consistent solution for GTV segmentation and shows potentials to be generally applied for radiation therapy planning of a variety of cancer (e.g. lung, sarcoma, liver and so on).

Automatic multiatlas based organ at risk segmentation in mice.

During the treatment planning of a preclinical small animal irradiation, which has time limitations for reasons of animal wellbeing and workflow efficiency, the time consuming organ at risk (OAR) delineation is performed manually. This work aimed to develop, demonstrate, and quantitatively evaluate an automated contouring method for six OARs in a preclinical irritation treatment workflow. Microconebeam CT images of nine healthy mice were contoured with an in-house developed multiatlas-based image segmentation (MABIS) algorithm for six OARs: kidneys, eyes, heart, and brain. The automatic contouring was compared with the manual delineation using three quantitative metrics: the Dice Similarity Coefficient (DSC), 95th percentile Hausdorff Distance, and the centre of mass displacement. A good agreement between manual and automatic contouring was found for OARs with sharp organ boundaries. For the brain and the heart, the median DSC was larger than 0.94, the median 95th Hausdorff Distance smaller than 0.44 mm, and the median centre of mass displacementsmaller than 0.20 mm. Lower DSC values were obtained for the other OARs, but the median DSC was still larger than 0.74 for the left eye, 0.69 for the right eye, 0.89 for the left kidney and 0.80 for the right kidney. The MABIS algorithm was able to delineate six OARs with a relatively high accuracy. Segmenting OARs with sharp organ boundaries performed better than low contrast OARs. A MABIS algorithm is developed, evaluated, and demonstrated in a preclinical small animal irradiation research workflow.

Quality assurance of radiotherapy in the ongoing EORTC 1219-DAHANCA-29 trial for HPV/p16 negative squamous cell carcinoma of the head and neck: Results of the benchmark case procedure

Background and purposeThe phase III EORTC 1219-DAHANCA 29 intergroup trial evaluates the influence of nimorazole in patients with locally advanced head and neck cancer when treated with accelerated radiotherapy (RT) in combination with chemotherapy. This article describes the results of the RT Benchmark Case (BC) performed before patient inclusion. Materials and methodsThe participating centers were asked to perform a 2-step BC, consisting of (1) a delineation and (2) a planning exercise according to the protocol guidelines. Submissions were prospectively centrally reviewed and feedback was given to the submitting centers. Sørensen–Dice similarity index (DSI) and the 95th percentile Hausdorff distance (HD) were retrospectively used to evaluate the agreement between the centers and the expert contours. ResultsFifty-four submissions (34 delineation and 20 planning exercises) from 19 centers were reviewed. Nine (47%) centers needed to perform the delineation step twice and three (16%) centers 3 times before receiving an approval. An increase in DSI-value and a decrease in HD, in particular for the prophylactic Clinical Target Volume (pCTV), could be found for the resubmitted cases. No unacceptable variations could be found for the planning exercise. ConclusionsThese BC-results highlight the need for effective and prospective RTQA in clinical trials. Even with clearly defined protocol guidelines, delineation and not planning remain the main reason for unacceptable protocol variations. The introduction of more objective quantitative analysis methods, such as the HD and DSI, in future trials might strengthen the evaluation by experts.

Multimodality Non-rigid Image Registration for Planning, Targeting and Monitoring During CT-Guided Percutaneous Liver Tumor Cryoablation

The aim of this study was to develop non-rigid image registration between preprocedure contrast-enhanced magnetic resonance (MR) images and intraprocedure unenhanced computed tomographic (CT) images, to enhance tumor visualization and localization during CT imaging-guided liver tumor cryoablation procedures. A non-rigid registration technique was evaluated with different preprocessing steps and algorithm parameters and compared to a standard rigid registration approach. The Dice similarity coefficient, target registration error, 95th-percentile Hausdorff distance, and total registration time (minutes) were compared using a two-sided Student's t test. The entire registration method was then applied during five CT imaging-guided liver cryoablation cases with the intraprocedural CT data transmitted directly from the CT scanner, with both accuracy and registration time evaluated. Selected optimal parameters for registration were a section thickness of 5 mm, cropping the field of view to 66% of its original size, manual segmentation of the liver, B-spline control grid of 5 × 5 × 5, and spatial sampling of 50,000 pixels. A mean 95th-percentile Hausdorff distance of 3.3 mm (a 2.5 times improvement compared to rigid registration, P < .05), a mean Dice similarity coefficient of 0.97 (a 13% increase), and a mean target registration error of 4.1 mm (a 2.7 times reduction) were measured. During the cryoablation procedure, registration between the preprocedure MR and the planning intraprocedure CT imaging took a mean time of 10.6 minutes, MR to targeting CT image took 4 minutes, and MR to monitoring CT imaging took 4.3 minutes. Mean registration accuracy was <3.4 mm. Non-rigid registration allowed improved visualization of the tumor during interventional planning, targeting, and evaluation of tumor coverage by the ice ball. Future work is focused on reducing segmentation time to make the method more clinically acceptable.