Virtual pretreatment patient-specific quality assurance of volumetric modulated arc therapy using deep learning

BACKGROUND

Automatic patient-specific quality assurance (PSQA) is recently explored using artificial intelligence approaches, and several studies reported the development of machine learning models for predicting the gamma pass rate (GPR) index only.

PURPOSE

To develop a novel deep learning approach using a generative adversarial network (GAN) to predict the synthetic measured fluence.

METHODS AND MATERIALS

A novel training method called "dual training," which involves the training of the encoder and decoder separately, was proposed and evaluated for cycle GAN (cycle-GAN) and conditional GAN (c-GAN). A total of 164 VMAT treatment plans, including 344 arcs (training data: 262, validation data: 30, and testing data: 52) from various treatment sites, were selected for prediction model development. For each patient, portal-dose-image-prediction fluence from TPS was used as input, and measured fluence from EPID was used as output/response for model training. Predicted GPR was derived by comparing the TPS fluence with the synthetic measured fluence generated by the DL models using gamma evaluation of criteria 2%/2 mm. The performance of dual training was compared against the traditional single-training approach. In addition, we also developed a separate classification model specifically designed to detect automatically three types of errors (rotational, translational, and MU-scale) in the synthetic EPID-measured fluence.

RESULTS

Overall, the dual training improved the prediction accuracy of both cycle-GAN and c-GAN. Predicted GPR results of single training were within 3% for 71.2% and 78.8% of test cases for cycle-GAN and c-GAN, respectively. Moreover, similar results for dual training were 82.7% and 88.5% for cycle-GAN and c-GAN, respectively. The error detection model showed high classification accuracy (>98%) for detecting errors related to rotational and translational errors. However, it struggled to differentiate the fluences with "MU scale error" from "error-free" fluences.

CONCLUSION

We developed a method to automatically generate the synthetic measured fluence and identify errors within them. The proposed dual training improved the PSQA prediction accuracy of both the GAN models, with c-GAN demonstrating superior performance over the cycle-GAN. Our results indicate that the c-GAN with dual training approach combined with error detection model, can accurately generate the synthetic measured fluence for VMAT PSQA and identify the errors. This approach has the potential to pave the way for virtual patient-specific QA of VMAT treatments.

Deep Learning Approaches for Automatic Quality Assurance of Magnetic Resonance Images Using ACR Phantom

BACKGROUND

In recent years, there has been a growing trend towards utilizing Artificial Intelligence (AI) and machine learning techniques in medical imaging, including for the purpose of automating quality assurance. In this research, we aimed to develop and evaluate various deep learning-based approaches for automatic quality assurance of Magnetic Resonance (MR) images using the American College of Radiology (ACR) standards.

METHODS

The study involved the development, optimization, and testing of custom convolutional neural network (CNN) models. Additionally, popular pre-trained models such as VGG16, VGG19, ResNet50, InceptionV3, EfficientNetB0, and EfficientNetB5 were trained and tested. The use of pre-trained models, particularly those trained on the ImageNet dataset, for transfer learning was also explored. Two-class classification models were employed for assessing spatial resolution and geometric distortion, while an approach classifying the image into 10 classes representing the number of visible spokes was used for the low contrast.

RESULTS

Our results showed that deep learning-based methods can be effectively used for MR image quality assurance and can improve the performance of these models. The low contrast test was one of the most challenging tests within the ACR phantom.

CONCLUSIONS

Overall, for geometric distortion and spatial resolution, all of the deep learning models tested produced prediction accuracy of 80% or higher. The study also revealed that training the models from scratch performed slightly better compared to transfer learning. For the low contrast, our investigation emphasized the adaptability and potential of deep learning models. The custom CNN models excelled in predicting the number of visible spokes, achieving commendable accuracy, recall, precision, and F1 scores.

Generating synthetic images from cone beam computed tomography using self-attention residual UNet for head and neck radiotherapy

Background and purpose

Accurate CT numbers in Cone Beam CT (CBCT) are crucial for precise dose calculations in adaptive radiotherapy (ART). This study aimed to generate synthetic CT (sCT) from CBCT using deep learning (DL) models in head and neck (HN) radiotherapy.

Materials and methods

A novel DL model, the 'self-attention-residual-UNet' (ResUNet), was developed for accurate sCT generation. ResUNet incorporates a self-attention mechanism in its long skip connections to enhance information transfer between the encoder and decoder. Data from 93 HN patients, each with planning CT (pCT) and first-day CBCT images were used. Model performance was evaluated using two DL approaches (non-adversarial and adversarial training) and two model types (2D axial only vs. 2.5D axial, sagittal, and coronal). ResUNet was compared with the traditional UNet through image quality assessment (Mean Absolute Error (MAE), Peak-Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM)) and dose calculation accuracy evaluation (DVH deviation and gamma evaluation (1 %/1mm)).

Results

Image similarity evaluation results for the 2.5D-ResUNet and 2.5D-UNet models were: MAE: 46±7 HU vs. 51±9 HU, PSNR: 66.6±2.0 dB vs. 65.8±1.8 dB, and SSIM: 0.81±0.04 vs. 0.79±0.05. There were no significant differences in dose calculation accuracy between DL models. Both models demonstrated DVH deviation below 0.5 % and a gamma-pass-rate (1 %/1mm) exceeding 97 %.

Conclusions

ResUNet enhanced CT number accuracy and image quality of sCT and outperformed UNet in sCT generation from CBCT. This method holds promise for generating precise sCT for HN ART.

Generation of synthetic CT from CBCT using deep learning approaches for head and neck cancer patients

Purpose.To create a synthetic CT (sCT) from daily CBCT using either deep residual U-Net (DRUnet), or conditional generative adversarial network (cGAN) for adaptive radiotherapy planning (ART).Methods.First fraction CBCT and planning CT (pCT) were collected from 93 Head and Neck patients who underwent external beam radiotherapy. The dataset was divided into training, validation, and test sets of 58, 10 and 25 patients respectively. Three methods were used to generate sCT, 1. Nonlocal means patch based method was modified to include multiscale patches defining the multiscale patch based method (MPBM), 2. An encoder decoder 2D Unet with imbricated deep residual units was implemented, 3. DRUnet was integrated to the generator part of cGAN whereas a convolutional PatchGAN classifier was used as the discriminator. The accuracy of sCT was evaluated geometrically using Mean Absolute Error (MAE). Clinical Volumetric Modulated Arc Therapy (VMAT) plans were copied from pCT to registered CBCT and sCT and dosimetric analysis was performed by comparing Dose Volume Histogram (DVH) parameters of planning target volumes (PTVs) and organs at risk (OARs). Furthermore, 3D Gamma analysis (2%/2mm, global) between the dose on the sCT or CBCT and that on the pCT was performed.Results. The average MAE calculated between pCT and CBCT was 180.82 ± 27.37HU. Overall, all approaches significantly reduced the uncertainties in CBCT. Deep learning approaches outperformed patch-based methods with MAE = 67.88 ± 8.39HU (DRUnet) and MAE = 72.52 ± 8.43HU (cGAN) compared to MAE = 90.69 ± 14.3HU (MPBM). The percentages of DVH metric deviations were below 0.55% for PTVs and 1.17% for OARs using DRUnet. The average Gamma pass rate was 99.45 ± 1.86% for sCT generated using DRUnet.Conclusion.DL approaches outperformed MPBM. Specifically, DRUnet could be used for the generation of sCT with accurate intensities and realistic description of patient anatomy. This could be beneficial for CBCT based ART.

Investigation of radiomics and deep convolutional neural networks approaches for glioma grading

Purpose.To determine glioma grading by applying radiomic analysis or deep convolutional neural networks (DCNN) and to benchmark both approaches on broader validation sets.Methods.Seven public datasets were considered: (1) low-grade glioma or high-grade glioma (369 patients, BraTS'20) (2) well-differentiated liposarcoma or lipoma (115, LIPO); (3) desmoid-type fibromatosis or extremity soft-tissue sarcomas (203, Desmoid); (4) primary solid liver tumors, either malignant or benign (186, LIVER); (5) gastrointestinal stromal tumors (GISTs) or intra-abdominal gastrointestinal tumors radiologically resembling GISTs (246, GIST); (6) colorectal liver metastases (77, CRLM); and (7) lung metastases of metastatic melanoma (103, Melanoma). Radiomic analysis was performed on 464 (2016) radiomic features for the BraTS'20 (others) datasets respectively. Random forests (RF), Extreme Gradient Boosting (XGBOOST) and a voting algorithm comprising both classifiers were tested. The parameters of the classifiers were optimized using a repeated nested stratified cross-validation process. The feature importance of each classifier was computed using the Gini index or permutation feature importance. DCNN was performed on 2D axial and sagittal slices encompassing the tumor. A balanced database was created, when necessary, using smart slices selection. ResNet50, Xception, EficientNetB0, and EfficientNetB3 were transferred from the ImageNet application to the tumor classification and were fine-tuned. Five-fold stratified cross-validation was performed to evaluate the models. The classification performance of the models was measured using multiple indices including area under the receiver operating characteristic curve (AUC).Results.The best radiomic approach was based on XGBOOST for all datasets; AUC was 0.934 (BraTS'20), 0.86 (LIPO), 0.73 (LIVER), (0.844) Desmoid, 0.76 (GIST), 0.664 (CRLM), and 0.577 (Melanoma) respectively. The best DCNN was based on EfficientNetB0; AUC was 0.99 (BraTS'20), 0.982 (LIPO), 0.977 (LIVER), (0.961) Desmoid, 0.926 (GIST), 0.901 (CRLM), and 0.89 (Melanoma) respectively.Conclusion.Tumor classification can be accurately determined by adapting state-of-the-art machine learning algorithms to the medical context.

Predicting respiratory motion using a novel patient specific dual deep recurrent neural networks

Real-time tracking of a target volume is a promising solution for reducing the planning margins and both dosimetric and geometric uncertainties in the treatment of thoracic and upper-abdomen cancers. Respiratory motion prediction is an integral part of real-time tracking to compensate for the latency of tracking systems. The purpose of this work was to develop a novel method for accurate respiratory motion prediction using dual deep recurrent neural networks (RNNs). The respiratory motion data of 111 patients were used to train and evaluate the method. For each patient, two models (Network1 and Network2) were trained on 80% of the respiratory wave, and the remaining 20% was used for evaluation. The first network (Network 1) is a "coarse resolution" prediction of future points and second network (Network 2) provides a "fine resolution" prediction to interpolate between the future predictions. The performance of the method was tested using two types of RNN algorithms : Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The accuracy of each model was evaluated using the root mean square error (RMSE) and mean absolute error (MAE). Overall, the RNN model with GRU- function had better accuracy than the RNN model with LSTM-function (RMSE (mm): 0.4±0.2 vs. 0.6±0.3; MAE (mm): 0.4±0.2 vs. 0.6±0.2). The GRU was able to predict the respiratory motion accurately (<1 mm) up to the latency period of 440 ms, and LSTM's accuracy was acceptable only up to 240 ms. The proposed method using GRU function can be used for respiratory-motion prediction up to a latency period of 440 ms.

Design and construction of a customizable phantom for the characterization of the three-dimensional magnetic resonance imaging geometric distortion

One of the main challenges to using magnetic resonance imaging (MRI) in radiotherapy is the existence of system‐related geometric inaccuracies caused mainly by the inhomogeneity in the main magnetic field and the nonlinearities of the gradient coils.

Several physical phantoms, with fixed configuration, have been developed and commercialized for the assessment of the MRI geometric distortion. In this study, we propose a new design of a customizable phantom that can fit any type of radio frequency (RF) coil. It is composed of 3D printed plastic blocks containing holes that can hold glass tubes which can be filled with any liquid. The blocks can be assembled to construct phantoms with any dimension.

The feasibility of this design has been demonstrated by assembling four phantoms with high robustness allowing the assessment of the geometric distortion for the GE split head coil, the head and neck array coil, the anterior array coil, and the body coil. Phantom reproducibility was evaluated by analyzing the geometric distortion on CT acquisition of five independent assemblages of the phantom.

This solution meets all expectations in terms of having a robust, lightweight, modular, and practical tool for measuring distortion in three dimensions. Mean error in the position of the tubes was less than 0.2 mm. For the geometric distortion, our results showed that for all typical MRI sequences used for radiotherapy, the mean geometric distortion was less than 1 mm and less than 2.5 mm over radial distances of 150 mm and 250 mm, respectively.

These tools will be part of a quality assurance program aimed at monitoring the image quality of MRI scanners used to guide radiation therapy.

SU-E-T-649: Evaluation of RapidArc- Based Stereotactic Cranial Radiotherapy Plans with MU Objective Using Multiple Non Coplanar Arcs in Comparison with Conventional Dynamic Conformal Arc Technique

PURPOSE

Previous researches reported that RapidArc plans for stereotactic cranial radiotherapy have two to three times more MUs as compared to Conventional Dynamic Conformal Arc (DCA) Technique. This study aims to evaluate RapidArc plans using multiple non- coplanar arcs, developed with MU objective constraint in the optimization stage.

METHODS

Five single brain metastasis and three multiple metastases cases previously planned using DCA techniques in BrainLab iPlan Version 4.1 were investigated in this study. For each case, the target was defined on CT-MR fused images in iPlan. The CT images and contours of these patients were exported from iPlan to Varian Eclipse TPS Version 8.6. For each case, a DCA plan and a RapidArc plan with multiple non-coplanar arcs with and without using MU objective in the optimization stage were generated using Varian Trilogy machine with Millennium 120 MLC keeping the same prescription and critical structure dose limits. All plans were evaluated according to Conformity Index (CI-modified Paddick) Homogeneity Index (HI), and the normal tissue volume receiving various dose levels (V80%, V50%, V25% and V10%).

RESULTS

In all the plans, the target objectives were met and dose to OARs was within tolerance dose constraints. RapidArc plans with and without MU objective showed better CI and HI as supposed to DCA plans. V80%, V50%, V25% and V10% of normal tissue for RapidArc plans are equal or lesser than DCA plans. Single isocentre RapidArc plan for closely spaced multiple metastases cases showed better dose fall off between the lesions as supposed to DCA plans. RapidArc plans with MU objective resulted in comparable MUs as that of DCA plans.

CONCLUSIONS

Our study showed RapidArc plans done with and without MU objective have no significant dosimetric difference in plan objectives. Therefore, multiple non-coplanar RapidArc plans with MU objective is clinically feasible and can provide better treatment plans than conventional DCA plans, especially for complicated cases.