Ensemble learning and personalized training for the improvement of unsupervised deep learning-based synthetic CT reconstruction.

BACKGROUND: The growing adoption of magnetic resonance imaging (MRI)-guided radiation therapy (RT) platforms and a focus on MRI-only RT workflows have brought the technical challenge of synthetic computed tomography (sCT) reconstruction to the forefront. Unpaired-data deep learning-based approaches to the problem offer the attractive characteristic of not requiring paired training data, but the gap between paired- and unpaired-data results can be limiting. PURPOSE: We present two distinct approaches aimed at improving unpaired-data sCT reconstruction results: a cascade ensemble that combines multiple models and a personalized training strategy originally designed for the paired-data setting. METHODS: Comparisons are made between the following models: (1) the paired-data fully convolutional DenseNet (FCDN), (2) the FCDN with the Intentional Deep Overfit Learning (IDOL) personalized training strategy, (3) the unpaired-data CycleGAN, (4) the CycleGAN with the IDOL training strategy, and (5) the CycleGAN as an intermediate model in a cascade ensemble approach. Evaluation of the various models over 25 total patients is carried out using a five-fold cross-validation scheme, with the patient-specific IDOL models being trained for the five patients of fold 3, chosen at random. RESULTS: In both the paired- and unpaired-data settings, adopting the IDOL training strategy led to improvements in the mean absolute error (MAE) between true CT images and sCT outputs within the body contour (mean improvement, paired- and unpaired-data approaches, respectively: 38%, 9%) and in regions of bone (52%, 5%), the peak signal-to-noise ratio (PSNR; 15%, 7%), and the structural similarity index (SSIM; 6%, <1%). The ensemble approach offered additional benefits over the IDOL approach in all three metrics (mean improvement over unpaired-data approach in fold 3; MAE: 20%; bone MAE: 16%; PSNR: 10%; SSIM: 2%), and differences in body MAE between the ensemble approach and the paired-data approach are statistically insignificant. CONCLUSIONS: We have demonstrated that both a cascade ensemble approach and a personalized training strategy designed initially for the paired-data setting offer significant improvements in image quality metrics for the unpaired-data sCT reconstruction task. Closing the gap between paired- and unpaired-data approaches is a step toward fully enabling these powerful and attractive unpaired-data frameworks.

Early Experience of Online Adaptive Radiation Therapy for Definitive Radiation of Patients With Head and Neck Cancer.

Purpose: The advent of cone beam computed tomography-based online adaptive radiation therapy (oART) has dramatically reduced the barriers of adaptation. We present the first prospective oART experience data in radiation of head and neck cancers (HNC). Methods and Materials: Patients with HNC receiving definitive standard fractionation (chemo)radiation who underwent at least 1 oART session were enrolled in a prospective registry study. The frequency of adaptations was at the discretion of the treating physician. Physicians were given the option of delivering 1 of 2 plans during adaptation: the original radiation plan transposed onto the cone beam computed tomography with adapted contours (scheduled), and a new adapted plan generated from the updated contours (adapted). A paired t test was used to compare the mean doses between scheduled and adapted plans. Results: Twenty-one patients (15 oropharynx, 4 larynx/hypopharynx, 2 other) underwent 43 adaptation sessions (median, 2). The median ART process time was 23 minutes, median physician time at the console was 27 minutes, and median patient time in the vault was 43.5 minutes. The adapted plan was chosen 93% of the time. The mean volume in each planned target volume (PTV) receiving 100% of the prescription dose for the scheduled versus adapted plan for high-risk PTVs was 87.8% versus 95% (P < .01), intermediate-risk PTVs was 87.3% versus 97.9% (P < .01), and low-risk PTVs was 94% versus 97.8% (P < .01), respectively. The mean hotspot was also lower with adaptation: 108.8% versus 106.4% (P < .01). All but 1 organ at risk (11/12) saw a decrease in their dose with the adapted plans, with the mean ipsilateral parotid (P = .013), mean larynx (P < .01), maximum point spinal cord (P < .01), and maximum point brain stem (P = .035) reaching statistical significance. Conclusions: Online ART is feasible for HNC, with significant improvement in target coverage and homogeneity and a modest decrease in doses to several organs at risk.

Acute coronary event (ACE) prediction following breast radiotherapy by features extracted from 3D CT, dose, and cardiac structures.

PURPOSE: Heart toxicity, such as major acute coronary events (ACE), following breast radiation therapy (RT) is of utmost concern. Thus, many studies have been investigating the effect of mean heart dose (MHD) and dose received in heart sub-structures on toxicity. Most studies focused on the dose thresholds in the heart and its sub-structures, while few studies adopted such computational methods as deep neural networks (DNN) and radiomics. This work aims to construct a feature-driven predictive model for ACE after breast RT. METHODS: A recently proposed two-step predictive model that extracts a number of features from a deep auto-segmentation network and processes the selected features for prediction was adopted. This work refined the auto-segmenting network and feature processing algorithms to enhance performance in cardiac toxicity prediction. In the predictive model, the deep convolutional neural network (CNN) extracted features from 3D computed tomography (CT) images and dose distributions in three automatically segmented heart sub-structures, including the left anterior descending artery (LAD), right coronary artery (RCA), and left ventricle (LV). The optimal feature processing workflow for the extracted features was explored to enhance the prediction accuracy. The regions associated with toxicity were visualized using a class activation map (CAM)-based technique. Our proposed model was validated against a conventional DNN (convolutional and fully connected layers) and radiomics with a patient cohort of 84 cases, including 29 and 55 patient cases with and without ACE. Of the entire 84 cases, 12 randomly chosen cases (5 toxicity and 7 non-toxicity cases) were set aside for independent test, and the remaining 72 cases were applied to 4-fold stratified cross-validation. RESULTS: Our predictive model outperformed the conventional DNN by 38% and 10% and radiomics-based predictive models by 9% and 10% in AUC for 4-fold cross-validations and independent test, respectively. The degree of enhancement was greater when incorporating dose information and heart sub-structures into feature extraction. The model whose inputs were CT, dose, and three sub-structures (LV, LAD, and RCA) reached 96% prediction accuracy on average and 0.94 area under the curve (AUC) on average in the cross-validation, and also achieved prediction accuracy of 83% and AUC of 0.83 in the independent test. On 10 correctly predicted cases out of 12 for the independent test, the activation maps implied that for cases of ACE toxicity, the higher intensity was more likely to be observed inside the LV. CONCLUSIONS: The proposed model characterized by modifications in model input with dose distributions and cardiac sub-structures, and serial processing of feature extraction and feature selection techniques can improve the predictive performance in ACE following breast RT.

Deep-Learning-Based Automatic Detection and Segmentation of Brain Metastases with Small Volume for Stereotactic Ablative Radiotherapy.

Recently, several efforts have been made to develop the deep learning (DL) algorithms for automatic detection and segmentation of brain metastases (BM). In this study, we developed an advanced DL model to BM detection and segmentation, especially for small-volume BM. From the institutional cancer registry, contrast-enhanced magnetic resonance images of 65 patients and 603 BM were collected to train and evaluate our DL model. Of the 65 patients, 12 patients with 58 BM were assigned to test-set for performance evaluation. Ground-truth for BM was assigned to one radiation oncologist to manually delineate BM and another one to cross-check. Unlike other previous studies, our study dealt with relatively small BM, so the area occupied by the BM in the high-resolution images were small. Our study applied training techniques such as the overlapping patch technique and 2.5-dimensional (2.5D) training to the well-known U-Net architecture to learn better in smaller BM. As a DL architecture, 2D U-Net was utilized by 2.5D training. For better efficacy and accuracy of a two-dimensional U-Net, we applied effective preprocessing include 2.5D overlapping patch technique. The sensitivity and average false positive rate were measured as detection performance, and their values were 97% and 1.25 per patient, respectively. The dice coefficient with dilation and 95% Hausdorff distance were measured as segmentation performance, and their values were 75% and 2.057 mm, respectively. Our DL model can detect and segment BM with small volume with good performance. Our model provides considerable benefit for clinicians with automatic detection and segmentation of BM for stereotactic ablative radiotherapy.

Clinical feasibility of deep learning-based auto-segmentation of target volumes and organs-at-risk in breast cancer patients after breast-conserving surgery.

BACKGROUND: In breast cancer patients receiving radiotherapy (RT), accurate target delineation and reduction of radiation doses to the nearby normal organs is important. However, manual clinical target volume (CTV) and organs-at-risk (OARs) segmentation for treatment planning increases physicians' workload and inter-physician variability considerably. In this study, we evaluated the potential benefits of deep learning-based auto-segmented contours by comparing them to manually delineated contours for breast cancer patients. METHODS: CTVs for bilateral breasts, regional lymph nodes, and OARs (including the heart, lungs, esophagus, spinal cord, and thyroid) were manually delineated on planning computed tomography scans of 111 breast cancer patients who received breast-conserving surgery. Subsequently, a two-stage convolutional neural network algorithm was used. Quantitative metrics, including the Dice similarity coefficient (DSC) and 95% Hausdorff distance, and qualitative scoring by two panels from 10 institutions were used for analysis. Inter-observer variability and delineation time were assessed; furthermore, dose-volume histograms and dosimetric parameters were also analyzed using another set of patient data. RESULTS: The correlation between the auto-segmented and manual contours was acceptable for OARs, with a mean DSC higher than 0.80 for all OARs. In addition, the CTVs showed favorable results, with mean DSCs higher than 0.70 for all breast and regional lymph node CTVs. Furthermore, qualitative subjective scoring showed that the results were acceptable for all CTVs and OARs, with a median score of at least 8 (possible range: 0-10) for (1) the differences between manual and auto-segmented contours and (2) the extent to which auto-segmentation would assist physicians in clinical practice. The differences in dosimetric parameters between the auto-segmented and manual contours were minimal. CONCLUSIONS: The feasibility of deep learning-based auto-segmentation in breast RT planning was demonstrated. Although deep learning-based auto-segmentation cannot be a substitute for radiation oncologists, it is a useful tool with excellent potential in assisting radiation oncologists in the future. Trial registration Retrospectively registered.