Search

Objective. To develop a deep ensemble learning (DEL) model with radiomics spatial encoding execution for improved glioma segmentation accuracy using multi-parametric magnetic resonance imaging (mp-MRI). Approach. This model was developed using 369 glioma patients with a four-modality mp-MRI protocol: T1, contrast-enhanced T1 (T1-Ce), T2, and FLAIR. In each modality volume, a 3D sliding kernel was implemented across the brain to capture image heterogeneity: 56 radiomic features were extracted within the kernel, resulting in a fourth-order tensor. Each radiomic feature can then be encoded as a 3D image volume, namely a radiomic feature map (RFM). For each patient, all RFMs extracted from all four modalities were processed using principal component analysis for dimension reduction, and the first four principal components (PCs) were selected. Next, a DEL model comprised of four U-Net sub-models was trained for the segmentation of a region-of-interest: each sub-model utilizes the mp-MRI and one of the four PCs as a five-channel input for 2D execution. Last, four softmax probability results given by the DEL model were superimposed and binarized using Otsu's method as the segmentation results. Three DEL models were trained to segment the enhancing tumor (ET), tumor core (TC), and whole tumor (WT), respectively. The segmentation results given by the proposed ensemble were compared to the mp-MRI-only U-Net results. Main Results. All three radiomics-incorporated DEL models were successfully implemented: compared to the mp-MRI-only U-net results, the dice coefficients of ET (0.777 → 0.817), TC (0.742 → 0.757), and WT (0.823 → 0.854) demonstrated improvement. The accuracy, sensitivity, and specificity results demonstrated similar patterns. Significance. The adopted radiomics spatial encoding execution enriches the image heterogeneity information that leads to the successful demonstration of the proposed DEL model, which offers a new tool for mp-MRI-based medical image segmentation., (© 2023 Institute of Physics and Engineering in Medicine.)

Objective . Deep learning (DL) models for fluence map prediction (FMP) have great potential to reduce treatment planning time in intensity-modulated radiation therapy (IMRT) by avoiding the lengthy inverse optimization process. This study aims to improve the rigor of input feature design in a DL-FMP model by examining how different designs of input features influence model prediction performance. Approach . This study included 231 head-and-neck intensity-modulated radiation therapy patients. Three input feature designs were investigated. The first design (D1) assumed that information of all critical structures from all beam angles should be combined to predict fluence maps. The second design (D2) assumed that local anatomical information was sufficient for predicting radiation intensity of a beamlet at a respective beam angle. The third design (D3) assumed the need for both local anatomical information and inter-beam modulation to predict radiation intensity values of the beamlets that intersect at a voxel. For each input design, we tailored the DL model accordingly. All models were trained using the same set of ground truth plans (GT plans). The plans generated by DL models (DL plans) were analyzed using key dose-volume metrics. One-way ANOVA with multiple comparisons correction (Bonferroni method) was performed (significance level = 0.05). Main results . For PTV-related metrics, all DL plans had significantly higher maximum dose ( p  < 0.001), conformity index ( p  < 0.001), and heterogeneity index ( p  < 0.001) compared to GT plans, with D2 being the worst performer. Meanwhile, except for cord+5 mm ( p  < 0.001), DL plans of all designs resulted in OAR dose metrics that are comparable to those of GT plans. Significance . Local anatomical information contains most of the information that DL models need to predict fluence maps for clinically acceptable OAR sparing. Input features from beam angles are needed to achieve the best PTV coverage. These results provide valuable insights for further improvement of DL-FMP models and DL models in general., (© 2022 Institute of Physics and Engineering in Medicine.)

Radiomics features extracted from medical images have been widely reported to be useful in the patient specific outcome modeling for variety of assessment and prediction purposes. Successful application of radiomics features as imaging biomarkers, however, is dependent on the robustness of the approach to the variation in each step of the modeling workflow. Variation in the input image quality is one of the main sources that impacts the reproducibility of radiomics analysis when a model is applied to broader range of medical imaging data. The quality of medical image is generally affected by both the scanner related factors such as image acquisition/reconstruction settings and the patient related factors such as patient motion. This article aimed to review the published literatures in this field that reported the impact of various imaging factors on the radiomics features through the change in image quality. The literatures were categorized by different imaging modalities and also tabulated based on the imaging parameters and the class of radiomics features included in the study. Strategies for image quality standardization were discussed based on the relevant literatures and recommendations for reducing the impact of image quality variation on the radiomics in multi-institutional clinical trial were summarized at the end of this article., (© 2022 Institute of Physics and Engineering in Medicine.)

Objective. 4D-CBCT provides phase-resolved images valuable for radiomics analysis for outcome prediction throughout treatment courses. However, 4D-CBCT suffers from streak artifacts caused by under-sampling, which severely degrades the accuracy of radiomic features. Previously we developed group-patient-trained deep learning methods to enhance the 4D-CBCT quality for radiomics analysis, which was not optimized for individual patients. In this study, a patient-specific model was developed to further improve the accuracy of 4D-CBCT based radiomics analysis for individual patients. Approach. This patient-specific model was trained with intra-patient data. Specifically, patient planning 4D-CT was augmented through image translation, rotation, and deformation to generate 305 CT volumes from 10 volumes to simulate possible patient positions during the onboard image acquisition. 72 projections were simulated from 4D-CT for each phase and were used to reconstruct 4D-CBCT using FDK back-projection algorithm. The patient-specific model was trained using these 305 paired sets of patient-specific 4D-CT and 4D-CBCT data to enhance the 4D-CBCT image to match with 4D-CT images as ground truth. For model testing, 4D-CBCT were simulated from a separate set of 4D-CT scan images acquired from the same patient and were then enhanced by this patient-specific model. Radiomics features were then extracted from the testing 4D-CT, 4D-CBCT, and enhanced 4D-CBCT image sets for comparison. The patient-specific model was tested using 4 lung-SBRT patients' data and compared with the performance of the group-based model. The impact of model dimensionality, region of interest (ROI) selection, and loss function on the model accuracy was also investigated. Main results. Compared with a group-based model, the patient-specific training model further improved the accuracy of radiomic features, especially for features with large errors in the group-based model. For example, the 3D whole-body and ROI loss-based patient-specific model reduces the errors of the first-order median feature by 83.67%, the wavelet LLL feature maximum by 91.98%, and the wavelet HLL skewness feature by 15.0% on average for the four patients tested. In addition, the patient-specific models with different dimensionality (2D versus 3D) or loss functions (L1 versus L1 + VGG + GAN) achieved comparable results for improving the radiomics accuracy. Using whole-body or whole-body+ROI L1 loss for the model achieved better results than using the ROI L1 loss alone as the loss function. Significance. This study demonstrated that the patient-specific model is more effective than the group-based model on improving the accuracy of the 4D-CBCT radiomic features analysis, which could potentially improve the precision for outcome prediction in radiotherapy., (© 2022 Institute of Physics and Engineering in Medicine.)

Objective: To design a deep transfer learning framework for modeling fluence map predictions for stereotactic body radiation therapy (SBRT) of adrenal cancer and similar sites that usually have a small number of cases. Approach: We developed a transfer learning framework for adrenal SBRT planning that leverages knowledge in a pancreas SBRT planning model. Treatment plans from the two sites had different dose prescriptions and beam settings but both prioritized gastrointestinal sparing. A base framework was first trained with 100 pancreas cases. This framework consists of two convolutional neural networks (CNN), which predict individual beam doses (BD-CNN) and fluence maps (FM-CNN) sequentially for 9-beam intensity-modulated radiation therapy (IMRT) plans. Forty-five adrenal plans were split into training/validation/test sets with the ratio of 20/10/15. The base BD-CNN was re-trained with transfer learning using 5/10/15/20 adrenal training cases to produce multiple candidate adrenal BD-CNN models. The base FM-CNN was directly used for adrenal cases. The deep learning (DL) plans were evaluated by several clinically relevant dosimetric endpoints, producing a percentage score relative to the clinical plans. Main results: Transfer learning significantly reduced the number of training cases and training time needed to train such a DL framework. The adrenal transfer learning model trained with 5/10/15/20 cases achieved validation plan scores of 85.4/91.2/90.7/89.4, suggesting that model performance saturated with 10 training cases. Meanwhile, a model using all 20 adrenal training cases without transfer learning only scored 80.5. For the final test set, the 5/10/15/20-case models achieved scores of 73.5/75.3/78.9/83.3. Significance: It is feasible to use deep transfer learning to train an IMRT fluence prediction framework. This technique could adapt to different dose prescriptions and beam configurations. This framework potentially enables DL modeling for clinical sites that have a limited dataset, either due to few cases or due to rapid technology evolution., (© 2021 Institute of Physics and Engineering in Medicine.)

Purpose. We have previously reported an artificial intelligence (AI) agent that automatically generates intensity-modulated radiation therapy (IMRT) plans via fluence map prediction, by-passing inverse planning. This AI agent achieved clinically comparable quality for prostate cases, but its performance on head-and-neck patients leaves room for improvement. This study aims to collect insights of the deep-learning-based (DL-based) fluence map prediction model by systematically analyzing its prediction errors. Methods. From the modeling perspective, the DL model's output is the fluence maps of IMRT plans. However, from the clinical planning perspective, the plan quality evaluation should be based on the clinical dosimetric criteria such as dose-volume histograms. To account for the complex and non-intuitive relationships between fluence map prediction errors and the corresponding dose distribution changes, we propose a novel error analysis approach that systematically examines plan dosimetric changes that are induced by varying amounts of fluence prediction errors. We investigated four decomposition modes of model prediction errors. The two spatial domain decompositions are based on fluence intensity and fluence gradient. The two frequency domain decompositions are based on Fourier-space banded frequency rings and Fourier-space truncated low-frequency disks. The decomposed error was analyzed for its impact on the resulting plans' dosimetric metrics. The analysis was conducted on 15 test cases spared from the 200 training and 16 validation cases used to train the model. Results. Most planning target volume metrics were significantly correlated with most error decompositions. The Fourier space disk radii had the largest Spearman's coefficients. The low-frequency region within a disk of ∼20% Fourier space contained most of errors that impact overall plan quality. Conclusions. This study demonstrates the feasibility of using fluence map prediction error analysis to understand the AI agent's performance. Such insights will help fine-tune the DL models in architecture design and loss function selection., (© 2021 The Author(s). Published on behalf of Institute of Physics and Engineering in Medicine by IOP Publishing Ltd.)

Objective . Synthesize realistic and controllable respiratory motions in the extended cardiac-torso (XCAT) phantoms by developing a generative adversarial network (GAN)-based deep learning technique. Methods . A motion generation model was developed using bicycle-GAN with a novel 4D generator. Input with the end-of-inhale (EOI) phase images and a Gaussian perturbation, the model generates inter-phase deformable-vector-fields (DVFs), which were composed and applied to the input to generate 4D images. The model was trained and validated using 71 4D-CT images from lung cancer patients and then applied to the XCAT EOI images to generate 4D-XCAT with realistic respiratory motions. A separate respiratory motion amplitude control model was built using decision tree regression to predict the input perturbation needed for a specific motion amplitude, and this model was developed using 300 4D-XCAT generated from 6 XCAT phantom sizes with 50 different perturbations for each size. In both patient and phantom studies, Dice coefficients for lungs and lung volume variation during respiration were compared between the simulated images and reference images. The generated DVF was evaluated by deformation energy. DVFs and ventilation maps of the simulated 4D-CT were compared with the reference 4D-CTs using cross correlation and Spearman's correlation. Comparison of DVFs and ventilation maps among the original 4D-XCAT, the generated 4D-XCAT, and reference patient 4D-CTs were made to show the improvement of motion realism by the model. The amplitude control error was calculated. Results . Comparing the simulated and reference 4D-CTs, the maximum deviation of lung volume during respiration was 5.8%, and the Dice coefficient reached at least 0.95 for lungs. The generated DVFs presented comparable deformation energy levels. The cross correlation of DVFs achieved 0.89 ± 0.10/0.86 ± 0.12/0.95 ± 0.04 along the x / y / z direction in the testing group. The cross correlation of ventilation maps derived achieved 0.80 ± 0.05/0.67 ± 0.09/0.68 ± 0.13, and the Spearman's correlation achieved 0.70 ± 0.05/0, 60 ± 0.09/0.53 ± 0.01, respectively, in the training/validation/testing groups. The generated 4D-XCAT phantoms presented similar deformation energy as patient data while maintained the lung volumes of the original XCAT phantom (Dice = 0.95, maximum lung volume variation = 4%). The motion amplitude control models controlled the motion amplitude control error to be less than 0.5 mm. Conclusions . The results demonstrated the feasibility of synthesizing realistic controllable respiratory motion in the XCAT phantom using the proposed method. This crucial development enhances the value of XCAT phantoms for various 4D imaging and therapy studies., (© 2021 Institute of Physics and Engineering in Medicine.)

Purpose: To investigate the impact of 4D-CBCT image quality on radiomic analysis and the efficacy of using deep learning based image enhancement to improve the accuracy of radiomic features of 4D-CBCT., Material and Methods: In this study, 4D-CT data from 16 lung cancer patients were obtained. Digitally reconstructed radiographs (DRRs) were simulated from the 4D-CT, and then used to reconstruct 4D CBCT using the conventional FDK (Feldkamp et al 1984 J. Opt. Soc. Am. A 1 612-9) algorithm. Different projection numbers (i.e. 72, 120, 144, 180) and projection angle distributions (i.e. evenly distributed and unevenly distributed using angles from real 4D-CBCT scans) were simulated to generate the corresponding 4D-CBCT. A deep learning model (TecoGAN) was trained on 10 patients and validated on 3 patients to enhance the 4D-CBCT image quality to match with the corresponding ground-truth 4D-CT. The remaining 3 patients with different tumor sizes were used for testing. The radiomic features in 6 different categories, including histogram, GLCM, GLRLM, GLSZM, NGTDM, and wavelet, were extracted from the gross tumor volumes of each phase of original 4D-CBCT, enhanced 4D-CBCT, and 4D-CT. The radiomic features in 4D-CT were used as the ground-truth to evaluate the errors of the radiomic features in the original 4D-CBCT and enhanced 4D-CBCT. Errors in the original 4D-CBCT demonstrated the impact of image quality on radiomic features. Comparison between errors in the original 4D-CBCT and enhanced 4D-CBCT demonstrated the efficacy of using deep learning to improve the radiomic feature accuracy., Results: 4D-CBCT image quality can substantially affect the accuracy of the radiomic features, and the degree of impact is feature-dependent. The deep learning model was able to enhance the anatomical details and edge information in the 4D-CBCT as well as removing other image artifacts. This enhancement of image quality resulted in reduced errors for most radiomic features. The average reduction of radiomics errors for 3 patients are 20.0%, 31.4%, 36.7%, 50.0%, 33.6% and 11.3% for histogram, GLCM, GLRLM, GLSZM, NGTDM and Wavelet features. And the error reduction was more significant for patients with larger tumors. The findings were consistent across different respiratory phases, projection numbers, and angle distributions., Conclusions: The study demonstrated that 4D-CBCT image quality has a significant impact on the radiomic analysis. The deep learning-based augmentation technique proved to be an effective approach to enhance 4D-CBCT image quality to improve the accuracy of radiomic analysis.

To investigate the feasibility of the log-demons deformable image registration (DIR) method to correct eddy current and field inhomogeneity distortions while preserving diffusion tensor information. Diffusion-weighted images (DWIs) are susceptible to distortions caused by eddy current and echo-planar imaging (EPI) gradients. We propose a post-acquisition correction algorithm using the log-demons DIR technique for eddy current and field inhomogeneity distortions of DWI. The new correction technique was applied to DWI acquired using a diffusion phantom and the multiple acquisitions for standardization of structural imaging validation and evaluation (MASSIVE) brain database. This method is compared to previous methods using cross-correlation, mutual information (MI). In the phantom study, the log-demons algorithm reduced eddy current and field inhomogeneity distortions while preserving diffusion tensor information when compared to affine and demon's registration techniques. Analysis of the tensor metrics using percent difference and the root mean square of the apparent diffusion coefficient and fractional anisotropy found that the log-demons algorithm outperforms the other algorithms in terms of preserving diffusion information. In the MASSIVE study, the average MI of all slices increased for both eddy current and field inhomogeneity distortion correction. The average absolute differences of all slices between corrected images with opposing gradients were also on average decreased. This work indicates that the log-demons DIR algorithm is feasible to reduce eddy current and field inhomogeneity distortions while preserving quantitative diffusion information.

Digital tomosynthesis (DTS) has been proposed as a fast low-dose imaging technique for image-guided radiation therapy (IGRT). However, due to the limited scanning angle, DTS reconstructed by the conventional FDK method suffers from significant distortions and poor plane-to-plane resolutions without full volumetric information, which severely limits its capability for image guidance. Although existing deep learning-based methods showed feasibilities in restoring volumetric information in DTS, they ignored the inter-patient variabilities by training the model using group patients. Consequently, the restored images still suffered from blurred and inaccurate edges. In this study, we presented a DTS enhancement method based on a patient-specific deep learning model to recover the volumetric information in DTS images. The main idea is to use the patient-specific prior knowledge to train the model to learn the patient-specific correlation between DTS and the ground truth volumetric images. To validate the performance of the proposed method, we enrolled both simulated and real on-board projections from lung cancer patient data. Results demonstrated the benefits of the proposed method: (1) qualitatively, DTS enhanced by the proposed method shows CT-like high image quality with accurate and clear edges; (2) quantitatively, the enhanced DTS has low-intensity errors and high structural similarity with respect to the ground truth CT images; (3) in the tumor localization study, compared to the ground truth CT-CBCT registration, the enhanced DTS shows 3D localization errors of ≤0.7 mm and ≤1.6 mm for studies using simulated and real projections, respectively; and (4), the DTS enhancement is nearly real-time. Overall, the proposed method is effective and efficient in enhancing DTS to make it a valuable tool for IGRT applications.

Purpose: To improve the prediction accuracy of respiratory signals by adapting the multi-layer perceptron neural network (MLP-NN) model to changing respiratory signals. We have previously developed an MLP-NN to predict respiratory signals obtained from a real-time position management (RPM) device. Preliminary testing results indicated that poor prediction accuracy may be observed after several seconds for irregular breathing patterns as only a set of fixed data was used in one-time training. To improve the prediction accuracy, we introduced a continuous learning technique using the updated training data to replace one-time learning using the fixed training data. We carried on this new prediction using an adaptation approach with dual MLP-NNs rather than single MLP-NN. When one MLP-NN was performing prediction of the respiratory signals, another one was being trained using the updated data and vice versa. The predicted performance was evaluated by root-mean-square-error (RMSE) between the predicted and true signals from 202 patients' respiratory patterns each with 1 min recording length. The effects of adding an additional network, training parameter, and respiratory signal irregularity on the performance of the new predictor were investigated based on four different network configurations: a single MLP-NN, high-computation dual MLP-NNs (U1), two different combinations of high- and low-computation dual MLP-NNs (U2 and U3). The RMSEs using U1 method were reduced by 34%, 19%, and 10% compared to those using MLP-NN, U2 and U3 methods, respectively. Continuous training of an MLP-NN based on a dual-network configuration using updated respiratory signals improved prediction accuracy compared to one-time training of an MLP-NN using fixed signals.

The purpose of this work was to develop a deep learning (DL) based algorithm, Automatic intensity-modulated radiotherapy (IMRT) Planning via Static Field Fluence Prediction (AIP-SFFP), for automated prostate IMRT planning with real-time planning efficiency. The following method was adopted: AIP-SFFP generates a prostate IMRT plan through predictions of fluence maps using patient anatomy. No inverse planning is required. AIP-SFFP is centered on a custom-built deep learning (DL) neural network for fluence map prediction. Predictions are imported to a commercial treatment-planning system for dose calculation and plan generation. AIP-SFFP was demonstrated for prostate IMRT simultaneously-integrated-boost planning (58.8 Gy/70 Gy to PTV 58.8 Gy /PTV 70 Gy in 28 fx, PTV = Planning Target Volume). Training data was generated from 106 patients using a knowledge-based planning (KBP) plan generator. Two types of 2D projection images were designed to represent structures' sizes and locations, and a total of eight projections were utilized to describe targets and organs-at-risk. Projections at nine template beam angles were stacked as inputs for artificial intelligence (AI) training. 14 patients were used as independent tests. The generated test plans were compared with the plans from the KBP training plan generator and clinic practice. The following results were obtained: After normalization (PTV 70 Gy V 70 Gy = 95%), all 14 AI plans met institutional criteria. The coverage of PTV 58.8 Gy in the AI plans was comparable to KBP and clinic plans without statistical significance. The whole body (BODY) D1cc and rectum D0.1cc of AI plans were slightly higher (<1 Gy) compared to KBP and clinic plans; in contrast, the bladder D1cc and other rectum and bladder low doses in the AI plans were slightly improved without clinical relevance. The overall isodose distribution in the AI plans was comparable with KBP plans and clinical plans. AIP-SFFP generated each test plan within 20s including the prediction and the dose calculation. In conclusion, AIP-SFFP was successfully developed for prostate IMRT planning. AIP-SFFP demonstrated good overall plan quality and real-time efficiency. Showing great promise, AIP-SFFP will be investigated for immediate clinical application.

Develop a machine learning-based method to generate multi-contrast anatomical textures in the 4D extended cardiac-torso (XCAT) phantom for more realistic imaging simulations. As a pilot study, we synthesize CT and CBCT textures in the chest region. For training purposes, major organs and gross tumor volumes (GTVs) in chest region were segmented from real patient images and assigned to different HU values to generate organ maps, which resemble the XCAT images. A dual-discriminator conditional-generative adversarial network (D-CGAN) was developed to synthesize anatomical textures in the corresponding organ maps. The D-CGAN was uniquely designed with two discriminators, one trained for the body and the other for the tumor. Various XCAT phantoms were input to the D-CGAN to generate textured XCAT phantoms. The D-CGAN model was trained separately using 62 CT and 63 CBCT images from lung SBRT patients to generate multi-contrast textured XCAT (MT-XCAT). The MT-XCAT phantoms were evaluated by comparing the intensity histograms and radiomic features with those from real patient images using Wilcoxon rank-sum test. The visual examination demonstrated that the MT-XCAT phantoms presented similar general contrast and anatomical textures as CT and CBCT images. The mean HU of the MT-XCAT-CT and MT-XCAT-CBCT were [Formula: see text] and [Formula: see text], compared with that of real CT ([Formula: see text]) and CBCT ([Formula: see text]). The majority of radiomic features from the MT-XCAT phantoms followed the same distribution as the real images according to the Wilcoxon rank-sum test, except for limited second-order features. The study demonstrated the feasibility of generating realistic MT-XCAT phantoms using D-CGAN. The MT-XCAT phantoms can be further expanded to include other modalities (MRI, PET, ultrasound, etc) under the same scheme. This crucial development greatly enhances the value of the phantom for various clinical applications, including testing and optimizing novel imaging techniques, validation of radiomics analysis methods, and virtual clinical trials.

To achieve accurate and fast deformable image registration (DIR) for pulmonary CT, we proposed a Multi-scale DIR framework with unsupervised Joint training of Convolutional Neural Network (MJ-CNN). MJ-CNN contains three models at multi-scale levels for a coarse-to-fine DIR to avoid being trapped in a local minimum. It is trained based on image similarity and deformation vector field (DVF) smoothness, requiring no supervision of ground-truth DVF. The three models are first trained sequentially and separately for their own registration tasks, and then are trained jointly for an end-to-end optimization under the multi-scale framework. In this study, MJ-CNN was trained using public SPARE 4D-CT data. The trained MJ-CNN was then evaluated on public DIR-LAB 4D-CT dataset as well as clinical CT-to-CBCT and CBCT-to-CBCT registration. For 4D-CT inter-phase registration, MJ-CNN achieved comparable accuracy to conventional iteration optimization-based methods, and showed the smallest registration errors compared to recently published deep learning-based DIR methods, demonstrating the efficacy of the proposed multi-scale joint training scheme. Besides, MJ-CNN trained using one dataset (SPARE) could generalize to a different dataset (DIR-LAB) acquired by different scanners and imaging protocols. Furthermore, MJ-CNN trained on 4D-CTs also performed well on CT-to-CBCT and CBCT-to-CBCT registration without any re-training or fine-tuning, demonstrating MJ-CNN's robustness against applications and imaging techniques. MJ-CNN took about 1.4 s for DVF estimation and required no manual-tuning of parameters during the evaluation. MJ-CNN is able to perform accurate DIR for pulmonary CT with nearly real-time speed, making it very applicable for clinical tasks.

To predict real-time 3D deformation field maps (DFMs) using Volumetric Cine MRI (VC-MRI) and adaptive boosting and multi-layer perceptron neural network (ADMLP-NN) for 4D target tracking. One phase of a prior 4D-MRI is set as the prior phase, MRI prior . Principal component analysis (PCA) is used to extract three major respiratory deformation modes from the DFMs generated between the prior and remaining phases. VC-MRI at each time-step is considered a deformation of MRI prior , where the DFM is represented as a weighted linear combination of the PCA components. The PCA weightings are solved by minimizing the differences between on-board 2D cine MRI and its corresponding VC-MRI slice. The PCA weightings solved during the initial training period are used to train an ADMLP-NN to predict PCA weightings ahead of time during the prediction period. The predicted PCA weightings are used to build predicted 3D DFM and ultimately, predicted VC-MRIs for 4D target tracking. The method was evaluated using a 4D computerized phantom (XCAT) with patient breathing curves and MRI data from a real liver cancer patient. Effects of breathing amplitude change and ADMLP-NN parameter variations were assessed. The accuracy of the PCA curve prediction was evaluated. The predicted real-time 3D tumor was evaluated against the ground-truth using volume dice coefficient (VDC), center-of-mass-shift (COMS), and target tracking errors. For the XCAT study, the average VDC and COMS for the predicted tumor were 0.92  ±  0.02 and 1.06  ±  0.40 mm, respectively, across all predicted time-steps. The correlation coefficients between predicted and actual PCA curves generated through VC-MRI estimation for the 1st/2nd principal components were 0.98/0.89 and 0.99/0.57 in the SI and AP directions, respectively. The optimal number of input neurons, hidden neurons, and MLP-NN for ADMLP-NN PCA weighting coefficient prediction were determined to be 7, 4, and 10, respectively. The optimal cost function threshold was determined to be 0.05. PCA weighting coefficient and VC-MRI accuracy was reduced for increased prediction-step size. Accurate PCA weighting coefficient prediction correlated with accurate VC-MRI prediction. For the patient study, the predicted 4D tumor tracking errors in superior-inferior, anterior-posterior and lateral directions were 0.50  ±  0.47 mm, 0.40  ±  0.55 mm, and 0.28  ±  0.12 mm, respectively. Preliminary studies demonstrated the feasibility to use VC-MRI and artificial neural networks to predict real-time 3D DFMs of the tumor for 4D target tracking.

The aim of this work was to develop a novel hybrid 3D hyperpolarized (HP) gas tagging MRI (t-MRI) technique and to evaluate it for lung respiratory motion measurement with comparison to deformable image registrations (DIR) methods. Three healthy subjects underwent a hybrid MRI which combines 3D HP gas t-MRI with a low resolution (Low-R, 4.5 mm isotropic voxels) 3D proton MRI (p-MRI), plus a high resolution (High-R, 2.5 mm isotropic voxels) 3D p-MRI, during breath-holds at the end-of-inhalation (EOI) and the end-of-exhalation (EOE). Displacement vector field (DVF) of the lung motion was determined from the t-MRI images by tracking tagging grids and from the High-R p-MRI using three DIR methods (B-spline based method implemented by Velocity, Free Form Deformation by MIM, and B-spline by an open source software Elastix: denoted as A, B, and C, respectively), labeled as tDVF and dDVF, respectively. The tDVF from the HP gas t-MRI was used as ground-truth reference to evaluate performance of the three DIR methods. Differences in both magnitude and angle between the tDVF and dDVFs were analyzed. The mean lung motion of the three subjects was 37.3 mm, 8.9 mm and 12.9 mm, respectively. Relatively large discrepancies were observed between the tDVF and the dDVFs as compared to previously reported DIR errors. The mean  ±  standard deviation (SD) DVF magnitude difference was 8.3  ±  5.6 mm, 9.2  ±  4.5 mm, and 9.3  ±  6.1 mm, and the mean  ±  SD DVF angular difference was 29.1  ±  12.1°, 50.1  ±  28.6°, and 39.0  ±  6.3°, for the DIR Methods A, B, and C, respectively. These preliminary results showed that the hybrid HP gas t-MRI technique revealed different lung motion patterns as compared to the DIR methods. It may provide unique perspectives in developing and evaluating DIR of the lungs. Novelty and Significance We designed a MRI protocol that includes a novel hybrid MRI technique (3D HP gas t-MRI with a low resolution 3D p-MRI) plus a high resolution 3D p-MRI. We tested the novel hybrid MRI technique on three healthy subjects for measuring regional lung respiratory motion with comparison to deformable image registrations (DIR) methods, and observed relatively large discrepancies in lung motion between HP gas t-MRI and DIR methods.

To clinically investigate the limited-projection CBCT (LP-CBCT) technology for daily positioning of patients receiving breath-hold lung SBRT radiation treatment and to investigate the feasibility of reconstructing fast 4D-CBCT from 1 min 3D-CBCT scan. Eleven patients who underwent breath-hold lung SBRT radiation treatment were scanned daily with on-board full-projection CBCT (CBCT) using half-fan scan. A subset of the CBCT projections and the prior planning CT were used to estimate the LP-CBCT images using the weighted free-form deformation method. The limited projections are clusteringly sampled within fifteen sub-angles in 360° in order to simulate the fast 1 min scan for 4D-CBCT. The estimated LP-CBCTs were rigidly registered to the planning CT to determine the clinical shifts needed for patient setup corrections, which were compared with shifts determined by the CBCT for evaluation. Both manual and automatic registrations were performed in order to compare the systematic registration errors. Fifty CBCT volumes were obtained from the eleven patients in fifty fractions for this pilot clinical study. For the CBCT images, the mean (±standard deviation) shifts between CBCT and planning CT from manual registration in left-right (LR), anterior-posterior (AP), and superior-inferior (SI) directions are 1.1  ±  1.2 mm, 2.1  ±  1.9 mm, 5.2  ±  3.6 mm, respectively. The mean deviation difference between shifts determined by CBCT and LP-CBCT images are 0.3  ±  0.5 mm, 0.5  ±  0.8 mm, 0.4  ±  0.3 mm, in LR, AP, and SI directions, respectively. The mean vector length of CBCT shift for all fractions is 6.1  ±  3.6 mm, and the mean vector length difference between CBCT and LP-CBCT for all fractions studied is 1.0  ±  0.9 mm. The automatic registrations yield similar results as manual registrations. The pilot clinical study shows that LP-CBCT localization offers comparable accuracy to CBCT localization for daily tumor positioning while reducing the projection number to 1/10 for patients receiving breath hold lung radiation treatment. The cluster projection sampling in this study also shows the feasibility of reconstructing fast 4D-CBCT from 1 min 3D-CBCT scan.

The purpose of this work was to investigate the potential relationship between radiomic features extracted from pre-treatment x-ray CT images and clinical outcomes following stereotactic body radiation therapy (SBRT) for non-small-cell lung cancer (NSCLC). Seventy patients who received SBRT for stage-1 NSCLC were retrospectively identified. The tumor was contoured on pre-treatment free-breathing CT images, from which 43 quantitative radiomic features were extracted to collectively capture tumor morphology, intensity, fine-texture, and coarse-texture. Treatment failure was defined based on cancer recurrence, local cancer recurrence, and non-local cancer recurrence following SBRT. The univariate association between each radiomic feature and each clinical endpoint was analyzed using Welch's t-test, and p-values were corrected for multiple hypothesis testing. Multivariate associations were based on regularized logistic regression with a singular value decomposition to reduce the dimensionality of the radiomics data. Two features demonstrated a statistically significant association with local failure: Homogeneity2 (p  =  0.022) and Long-Run-High-Gray-Level-Emphasis (p  =  0.048). These results indicate that relatively dense tumors with a homogenous coarse texture might be linked to higher rates of local recurrence. Multivariable logistic regression models produced maximum [Formula: see text] values of [Formula: see text], and [Formula: see text], for the recurrence, local recurrence, and non-local recurrence endpoints, respectively. The CT-based radiomic features used in this study may be more associated with local failure than non-local failure following SBRT for stage I NSCLC. This finding is supported by both univariate and multivariate analyses.

The purpose of this research was to study the sensitivity of Computed Tomography (CT) radiomic features to motion blurring and signal-to-noise ratios (SNR), and investigate its downstream effect regarding the classification of non-small cell lung cancer (NSCLC) histology. Forty-three radiomic features were considered and classified into one of four categories: Morphological, Intensity, Fine Texture, and Coarse Texture. First, a series of simulations were used to study feature-sensitivity to changes in spatial-temporal resolution. A dynamic digital phantom was used to generate images with different breathing amplitudes and SNR, from which features were extracted and characterized relative to initial simulation conditions. Stage I NSCLC patients were then retrospectively identified, from which three different acquisition-specific feature-spaces were generated based on free-breathing (FB), average-intensity-projection (AIP), and end-of-exhalation (EOE) CT images. These feature-spaces were derived to cover a wide range of spatial-temporal tradeoff. Normalized percent differences and concordance correlation coefficients (CCC) were used to assess the variability between the 3D and 4D acquisition techniques. Subsequently, three corresponding acquisition-specific logistic regression models were developed to classify lung tumor histology. Classification performance was compared between the different data-dependent models. Simulation results demonstrated strong linear dependences (p  >  0.95) between respiratory motion and morphological features, as well as between SNR and texture features. The feature Short Run Emphasis was found to be particularly stable to both motion blurring and changes in SNR. When comparing FB-to-EOE, 37% of features demonstrated high CCC agreement (CCC  >  0.8), compared to only 30% for FB-to-AIP. In classifying tumor histology, EoE images achieved an average AUC, Accuracy, Sensitivity, and Specificity of [Formula: see text], respectively. FB images achieved respective values of [Formula: see text], and AIP images achieved respective values of [Formula: see text]. Several radiomic features have been identified as being relatively robust to spatial-temporal variations based on both simulation data and patient data. In general, features that were sensitive to motion blurring were not necessarily the same features that were sensitive to changes in SNR. Our modeling results suggest that the EoE phase of a 4DCT acquisition may provide useful radiomic information, particularly for features that are highly sensitive to respiratory motion.

Beam angle configuration is a major planning decision in intensity modulated radiation treatment (IMRT) that has a significant impact on dose distributions and thus quality of treatment, especially in complex planning cases such as those for lung cancer treatment. We propose a novel method to automatically determine beam configurations that incorporates noncoplanar beams. We then present a completely automated IMRT planning algorithm that combines the proposed method with a previously reported OAR DVH prediction model. Finally, we validate this completely automatic planning algorithm using a set of challenging lung IMRT cases. A beam efficiency index map is constructed to guide the selection of beam angles. This index takes into account both the dose contributions from individual beams and the combined effect of multiple beams by introducing a beam-spread term. The effect of the beam-spread term on plan quality was studied systematically and the weight of the term to balance PTV dose conformity against OAR avoidance was determined. For validation, complex lung cases with clinical IMRT plans that required the use of one or more noncoplanar beams were re-planned with the proposed automatic planning algorithm. Important dose metrics for the PTV and OARs in the automatic plans were compared with those of the clinical plans. The results are very encouraging. The PTV dose conformity and homogeneity in the automatic plans improved significantly. And all the dose metrics of the automatic plans, except the lung V 5 Gy , were statistically better than or comparable with those of the clinical plans. In conclusion, the automatic planning algorithm can incorporate non-coplanar beam configurations in challenging lung cases and can generate plans efficiently with quality closely approximating that of clinical plans.

Purpose: compressed sensing reconstruction using total variation (TV) tends to over-smooth the edge information by uniformly penalizing the image gradient. The goal of this study is to develop a novel prior contour based TV (PCTV) method to enhance the edge information in compressed sensing reconstruction for CBCT., Methods: the edge information is extracted from prior planning-CT via edge detection. Prior CT is first registered with on-board CBCT reconstructed with TV method through rigid or deformable registration. The edge contours in prior-CT is then mapped to CBCT and used as the weight map for TV regularization to enhance edge information in CBCT reconstruction. The PCTV method was evaluated using extended-cardiac-torso (XCAT) phantom, physical CatPhan phantom and brain patient data. Results were compared with both TV and edge preserving TV (EPTV) methods which are commonly used for limited projection CBCT reconstruction. Relative error was used to calculate pixel value difference and edge cross correlation was defined as the similarity of edge information between reconstructed images and ground truth in the quantitative evaluation., Results: compared to TV and EPTV, PCTV enhanced the edge information of bone, lung vessels and tumor in XCAT reconstruction and complex bony structures in brain patient CBCT. In XCAT study using 45 half-fan CBCT projections, compared with ground truth, relative errors were 1.5%, 0.7% and 0.3% and edge cross correlations were 0.66, 0.72 and 0.78 for TV, EPTV and PCTV, respectively. PCTV is more robust to the projection number reduction. Edge enhancement was reduced slightly with noisy projections but PCTV was still superior to other methods. PCTV can maintain resolution while reducing the noise in the low mAs CatPhan reconstruction. Low contrast edges were preserved better with PCTV compared with TV and EPTV., Conclusion: PCTV preserved edge information as well as reduced streak artifacts and noise in low dose CBCT reconstruction. PCTV is superior to TV and EPTV methods in edge enhancement, which can potentially improve the localization accuracy in radiation therapy.

Purpose: To accelerate volumetric cine MRI (VC-MRI) using undersampled 2D-cine MRI to provide real-time 3D guidance for gating/target tracking in radiotherapy., Methods: 4D-MRI is acquired during patient simulation. One phase of the prior 4D-MRI is selected as the prior images, designated as MRI prior . The on-board VC-MRI at each time-step is considered a deformation of the MRI prior . The deformation field map is represented as a linear combination of the motion components extracted by principal component analysis from the prior 4D-MRI. The weighting coefficients of the motion components are solved by matching the corresponding 2D-slice of the VC-MRI with the on-board undersampled 2D-cine MRI acquired. Undersampled Cartesian and radial k-space acquisition strategies were investigated. The effects of k-space sampling percentage (SP) and distribution, tumor sizes and noise on the VC-MRI estimation were studied. The VC-MRI estimation was evaluated using XCAT simulation of lung cancer patients and data from liver cancer patients. Volume percent difference (VPD) and Center of Mass Shift (COMS) of the tumor volumes and tumor tracking errors were calculated., Results: For XCAT, VPD/COMS were 11.93  ±  2.37%/0.90  ±  0.27 mm and 11.53  ±  1.47%/0.85  ±  0.20 mm among all scenarios with Cartesian sampling (SP  =  10%) and radial sampling (21 spokes, SP  =  5.2%), respectively. When tumor size decreased, higher sampling rate achieved more accurate VC-MRI than lower sampling rate. VC-MRI was robust against noise levels up to SNR  =  20. For patient data, the tumor tracking errors in superior-inferior, anterior-posterior and lateral (LAT) directions were 0.46  ±  0.20 mm, 0.56  ±  0.17 mm and 0.23  ±  0.16 mm, respectively, for Cartesian-based sampling with SP  =  20% and 0.60  ±  0.19 mm, 0.56  ±  0.22 mm and 0.42  ±  0.15 mm, respectively, for radial-based sampling with SP  =  8% (32 spokes)., Conclusions: It is feasible to estimate VC-MRI from a single undersampled on-board 2D cine MRI. Phantom and patient studies showed that the temporal resolution of VC-MRI can potentially be improved by 5-10 times using a 2D cine image acquired with 10-20% k-space sampling.

Compared to other radiation therapy modalities, clinical electron beam therapy has remained practically unchanged for the past few decades even though electron beams with multiple energies are widely available on most linacs. In this paper, we present the concept of dynamic electron arc radiotherapy (DEAR), a new conformal electron therapy technique with synchronized couch motion. DEAR utilizes combination of gantry rotation, couch motion, and dose rate modulation to achieve desirable dose distributions in patient. The electron applicator is kept to minimize scatter and maintain narrow penumbra. The couch motion is synchronized with the gantry rotation to avoid collision between patient and the electron cone. In this study, we investigate the feasibility of DEAR delivery and demonstrate the potential of DEAR to improve dose distributions on simple cylindrical phantoms. DEAR was delivered on Varian's TrueBeam linac in Research Mode. In conjunction with the recorded trajectory log files, mechanical motion accuracies and dose rate modulation precision were analyzed. Experimental and calculated dose distributions were investigated for different energies (6 and 9 MeV) and cut-out sizes (1×10 cm(2) and 3×10 cm(2) for a 15×15 cm(2) applicator). Our findings show that DEAR delivery is feasible and has the potential to deliver radiation dose with high accuracy (root mean square error, or RMSE of <0.1 MU, <0.1° gantry, and <0.1 cm couch positions) and good dose rate precision (1.6 MU min(-1)). Dose homogeneity within ±2% in large and curved targets can be achieved while maintaining penumbra comparable to a standard electron beam on a flat surface. Further, DEAR does not require fabrication of patient-specific shields. These benefits make DEAR a promising technique for conformal radiotherapy of superficial tumors.

The aim of this study was to explore similarities between intensity-modulated radiotherapy (IMRT) and intensity-modulated arc therapy (IMAT) techniques in the context of the number of multi-leaf collimator (MLC) segments required to achieve plan objectives, the major factor influencing plan quality. Three clinical cases with increasing complexity were studied: (a) prostate only, (b) prostate and seminal vesicles and (c) prostate and pelvic lymph nodes. Initial 'gold-standard' plans with the maximum possible organ-at-risk sparing were generated for all three cases. For each case, multiple IMRT and IMAT plans were generated with varying intensity levels (IMRT) and arc control points (IMAT), which translate into varying MLC segments in both modalities. The IMAT/IMRT plans were forced to mimic the organ-at-risk sparing and target coverage in the gold-standard plans, thereby only allowing the target dose inhomogeneity to be variable. A higher target dose inhomogeneity (quantified as D5--dose to the highest 5% of target volume) implies that the plan is less capable of modulation. For each case, given a similar number of MLC segments, both IMRT and IMAT plans exhibit similar target dose inhomogeneity, indicating that there is no difference in their ability to provide dose painting. Target dose inhomogeneity remained approximately constant with decreasing segments, but sharply increased below a specific critical number of segments (70, 100, 110 for cases a, b, c, respectively). For the cases studied, IMAT and IMRT plans are similar in their dependence on the number of MLC segments. A minimum critical number of segments are required to ensure adequate plan quality. Future studies are needed to establish the range of minimum critical number of segments for different treatment sites and target-organ geometries.

This work investigates the use of the self-organizing map (SOM) technique for predicting lung radiation pneumonitis (RP) risk. SOM is an effective method for projecting and visualizing high-dimensional data in a low-dimensional space (map). By projecting patients with similar data (dose and non-dose factors) onto the same region of the map, commonalities in their outcomes can be visualized and categorized. Once built, the SOM may be used to predict pneumonitis risk by identifying the region of the map that is most similar to a patient's characteristics. Two SOM models were developed from a database of 219 lung cancer patients treated with radiation therapy (34 clinically diagnosed with Grade 2+ pneumonitis). The models were: SOM(all) built from all dose and non-dose factors and, for comparison, SOM(dose) built from dose factors alone. Both models were tested using ten-fold cross validation and Receiver Operating Characteristics (ROC) analysis. Models SOM(all) and SOM(dose) yielded ten-fold cross-validated ROC areas of 0.73 (sensitivity/specificity = 71%/68%) and 0.67 (sensitivity/specificity = 63%/66%), respectively. The significant difference between the cross-validated ROC areas of these two models (p < 0.05) implies that non-dose features add important information toward predicting RP risk. Among the input features selected by model SOM(all), the two with highest impact for increasing RP risk were: (a) higher mean lung dose and (b) chemotherapy prior to radiation therapy. The SOM model developed here may not be extrapolated to treatment techniques outside that used in our database, such as several-field lung intensity modulated radiation therapy or gated radiation therapy.