Cancer Control, Toxicity, and Secondary Malignancy Risks of Proton Radiation Therapy for Stage I-IIB Testicular Seminoma.

Purpose: This study's objective was to report cancer control and toxicity outcomes after proton radiation therapy (RT) in testicular seminoma and to compare secondary malignancy (SMN) risks with photon-based treatment alternatives. Methods and Materials: Consecutive patients with stage I-IIB testicular seminoma treated with proton RT at a single institution were retrospectively analyzed. Kaplan-Meier estimates for disease-free and overall survival were computed. Toxicities were scored using Common Terminology Criteria for Adverse Events version 5.0. Photon comparison plans, including 3-dimensional conformal RT (3D-CRT) and intensity modulated RT (IMRT)/volumetric arc therapy (VMAT), were created for each patient. Dosimetric parameters and SMN risk predictions for different in-field organs-at-risk were compared between the techniques. Excess absolute SMN risks were estimated with organ equivalent dose modeling. Results: Twenty-four patients were included (median age, 38.5 years). The majority of patients had stage II disease (IIA, 12 [50.0%]; IIB, 11 [45.8%]; IA, 1 [4.2%]). Seven (29.2%) and 17 (70.8%) patients had de novo and recurrent disease, respectively (de novo/recurrent: IA, 1/0; IIA, 4/8; IIB, 2/9). Most acute toxicities were mild (grade 1 [G1], 79.2%; G2, 12.5%) with G1 nausea being most common (70.8%). No serious events (G3-5) occurred. With a median follow-up time of 3 years (interquartile range, 2.1-3.6 years), 3-year disease-free and overall survival rates were 90.9% (95% confidence interval, 68.1%-97.6%) and 100% (95% confidence interval, 100%-100%), respectively. There were no documented late toxicities in the follow-up period, including worsening serial creatinine levels suggestive of early nephrotoxicity. Proton RT had significant reductions in mean organ-at-risk doses to the kidneys, stomach, colon, liver, bladder, and body compared with both 3D-CRT and IMRT/VMAT. Proton RT had significantly lower SMN risk predictions compared with 3D-CRT and IMRT/VMAT. Conclusions: Cancer control and toxicity outcomes using proton RT in stage I-IIB testicular seminoma are consistent with existing photon-based RT literature. However, proton RT may be associated with significantly lower SMN risks.

Real-time Markerless Tracking of Lung Tumors based on 2-D Fluoroscopy Imaging using Convolutional LSTM.

Purpose: To investigate the feasibility of tracking targets in 2D fluor images using a novel deep learning network. Methods: Our model design aims to capture the consistent motion of tumors in fluoroscopic images by neural network. Specifically, the model is trained by generative adversarial methods. The network is a coarse-to-fine architecture design. Convolutional LSTM (Long Short-term Memory) modules are introduced to account for the time correlation between different frames of the fluoroscopic images. The model was trained and tested on a digital X-CAT phantom in two studies. Series of coherent 2D fluoroscopic images representing the full respiration cycle were fed into the model to predict the localized tumor regions. In first study to test on massive scenarios, phantoms of different scales, tumor positions, sizes, and respiration amplitudes were generated to evaluate the accuracy of the model comprehensively. In second study to test on specific sample, phantoms were generated with fixed body and tumor sizes but different respiration amplitudes to investigate the effects of motion amplitude on the tracking accuracy. The tracking accuracy was quantitatively evaluated using intersection over union (IOU), tumor area difference, and centroid of mass difference (COMD). Results: In the first comprehensive study, the mean IOU and dice coefficient achieved 0.93±0.04 and 0.96±0.02. The mean tumor area difference was 4.34%±4.04%. And the COMD was 0.16 cm and 0.07 cm on average in SI (superior-interior) and LR (left-right) directions, respectively. In the second amplitude study, the mean IOU and dice coefficient achieved 0.98 and 0.99. The mean tumor difference was 0.17%. And the COMD was 0.03cm and 0.01 cm on average in SI and LR directions, respectively. Results demonstrated the robustness of our model against breathing variations. Conclusion: Our study showed the feasibility of using deep learning to track targets in x-ray fluoroscopic projection images without the aid of markers. The technique can be valuable for both pre- and during-treatment real-time target verification using fluoroscopic imaging in lung SBRT treatments.

Fast four-dimensional cone-beam computed tomography reconstruction using deformable convolutional networks.

BACKGROUND: Although four-dimensional cone-beam computed tomography (4D-CBCT) is valuable to provide onboard image guidance for radiotherapy of moving targets, it requires a long acquisition time to achieve sufficient image quality for target localization. To improve the utility, it is highly desirable to reduce the 4D-CBCT scanning time while maintaining high-quality images. Current motion-compensated methods are limited by slow speed and compensation errors due to the severe intraphase undersampling. PURPOSE: In this work, we aim to propose an alternative feature-compensated method to realize the fast 4D-CBCT with high-quality images. METHODS: We proposed a feature-compensated deformable convolutional network (FeaCo-DCN) to perform interphase compensation in the latent feature space, which has not been explored by previous studies. In FeaCo-DCN, encoding networks extract features from each phase, and then, features of other phases are deformed to those of the target phase via deformable convolutional networks. Finally, a decoding network combines and decodes features from all phases to yield high-quality images of the target phase. The proposed FeaCo-DCN was evaluated using lung cancer patient data. RESULTS: (1) FeaCo-DCN generated high-quality images with accurate and clear structures for a fast 4D-CBCT scan; (2) 4D-CBCT images reconstructed by FeaCo-DCN achieved 3D tumor localization accuracy within 2.5 mm; (3) image reconstruction is nearly real time; and (4) FeaCo-DCN achieved superior performance by all metrics compared to the top-ranked techniques in the AAPM SPARE Challenge. CONCLUSION: The proposed FeaCo-DCN is effective and efficient in reconstructing 4D-CBCT while reducing about 90% of the scanning time, which can be highly valuable for moving target localization in image-guided radiotherapy.

Quantification of lung function on CT images based on pulmonary radiomic filtering.

PURPOSE: To develop a radiomics filtering technique for characterizing spatial-encoded regional pulmonary ventilation information on lung computed tomography (CT). METHODS: The lung volume was segmented on 46 CT images, and a 3D sliding window kernel was implemented across the lung volume to capture the spatial-encoded image information. Fifty-three radiomic features were extracted within the kernel, resulting in a fourth-order tensor object. As such, each voxel coordinate of the original lung was represented as a 53-dimensional feature vector, such that radiomic features could be viewed as feature maps within the lungs. To test the technique as a potential pulmonary ventilation biomarker, the radiomic feature maps were compared to paired functional images (Galligas PET or DTPA-SPECT) based on the Spearman correlation (ρ) analysis. RESULTS: The radiomic feature maps GLRLM-based Run-Length Non-Uniformity and GLCOM-based Sum Average are found to be highly correlated with the functional imaging. The achieved ρ (median [range]) for the two features are 0.46 [0.05, 0.67] and 0.45 [0.21, 0.65] across 46 patients and 2 functional imaging modalities, respectively. CONCLUSIONS: The results provide evidence that local regions of sparsely encoded heterogeneous lung parenchyma on CT are associated with diminished radiotracer uptake and measured lung ventilation defects on PET/SPECT imaging. These findings demonstrate the potential of radiomics to serve as a complementary tool to the current lung quantification techniques and provide hypothesis-generating data for future studies.

Enhancement of 4-D Cone-Beam Computed Tomography (4D-CBCT) Using a Dual-Encoder Convolutional Neural Network (DeCNN).

4D-CBCT is a powerful tool to provide respiration-resolved images for the moving target localization. However, projections in each respiratory phase are intrinsically under-sampled under the clinical scanning time and imaging dose constraints. Images reconstructed by compressed sensing (CS)-based methods suffer from blurred edges. Introducing the average-4D-image constraint to the CS-based reconstruction, such as prior-image-constrained CS (PICCS), can improve the edge sharpness of the stable structures. However, PICCS can lead to motion artifacts in the moving regions. In this study, we proposed a dual-encoder convolutional neural network (DeCNN) to realize the average-image-constrained 4D-CBCT reconstruction. The proposed DeCNN has two parallel encoders to extract features from both the under-sampled target phase images and the average images. The features are then concatenated and fed into the decoder for the high-quality target phase image reconstruction. The reconstructed 4D-CBCT using of the proposed DeCNN from the real lung cancer patient data showed (1) qualitatively, clear and accurate edges for both stable and moving structures; (2) quantitatively, low-intensity errors, high peak signal-to-noise ratio, and high structural similarity compared to the ground truth images; and (3) superior quality to those reconstructed by several other state-of-the-art methods including the back-projection, CS total-variation, PICCS, and the single-encoder CNN. Overall, the proposed DeCNN is effective in exploiting the average-image constraint to improve the 4D-CBCT image quality.

Patient-specific deep learning model to enhance 4D-CBCT image for radiomics analysis.

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.

Building a patient-specific model using transfer learning for four-dimensional cone beam computed tomography augmentation.

BACKGROUND: We previously developed a deep learning model to augment the quality of four-dimensional (4D) cone-beam computed tomography (CBCT). However, the model was trained using group data, and thus was not optimized for individual patients. Consequently, the augmented images could not depict small anatomical structures, such as lung vessels. METHODS: In the present study, the transfer learning method was used to further improve the performance of the deep learning model for individual patients. Specifically, a U-Net-based model was first trained to augment 4D-CBCT using group data. Next, transfer learning was used to fine tune the model based on a specific patient's available data to improve its performance for that individual patient. Two types of transfer learning were studied: layer-freezing and whole-network fine-tuning. The performance of the transfer learning model was evaluated by comparing the augmented CBCT images with the ground truth images both qualitatively and quantitatively using a structure similarity index matrix (SSIM) and peak signal-to-noise ratio (PSNR). The results were also compared to those obtained using only the U-Net method. RESULTS: Qualitatively, the patient-specific model recovered more detailed information of the lung area than the group-based U-Net model. Quantitatively, the SSIM improved from 0.924 to 0.958, and the PSNR improved from 33.77 to 38.42 for the whole volumetric images for the group-based U-Net and patient-specific models, respectively. The layer-freezing method was found to be more efficient than the whole-network fine-tuning method, and had a training time as short as 10 minutes. The effect of augmentation by transfer learning increased as the number of projections used for CBCT reconstruction decreased. CONCLUSIONS: Overall, the patient-specific model optimized by transfer learning was efficient and effective at improving image qualities of augmented undersampled three-dimensional (3D)- and 4D-CBCT images, and could be extremely valuable for applications in image-guided radiation therapy.

4D radiomics: impact of 4D-CBCT image quality on radiomic analysis.

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.

Prior image-guided cone-beam computed tomography augmentation from under-sampled projections using a convolutional neural network.

BACKGROUND: Acquiring sparse-view cone-beam computed tomography (CBCT) is an effective way to reduce the imaging dose. However, images reconstructed by the conventional filtered back-projection method suffer from severe streak artifacts due to the projection under-sampling. Existing deep learning models have demonstrated feasibilities in restoring volumetric structures from the highly under-sampled images. However, because of the inter-patient variabilities, they failed to restore the patient-specific details with the common restoring pattern learned from the group data. Although the patient-specific models have been developed by training models using the intra-patient data and have shown effectiveness in restoring the patient-specific details, the models have to be retrained to be exclusive for each patient. It is highly desirable to develop a generalized model that can utilize the patient-specific information for the under-sampled image augmentation. METHODS: In this study, we proposed a merging-encoder convolutional neural network (MeCNN) to realize the prior image-guided under-sampled CBCT augmentation. Instead of learning the patient-specific structures, the proposed model learns a generalized pattern of utilizing the patient-specific information in the prior images to facilitate the under-sampled image enhancement. Specifically, the MeCNN consists of a merging-encoder and a decoder. The merging-encoder extracts image features from both the prior CT images and the under-sampled CBCT images, and merges the features at multi-scale levels via deep convolutions. The merged features are then connected to the decoders via shortcuts to yield high-quality CBCT images. The proposed model was tested on both the simulated CBCTs and the clinical CBCTs. The predicted CBCT images were evaluated qualitatively and quantitatively in terms of image quality and tumor localization accuracy. Mann-Whitney U test was conducted for the statistical analysis. P<0.05 was considered statistically significant. RESULTS: The proposed model yields CT-like high-quality CBCT images from only 36 half-fan projections. Compared to other methods, CBCT images augmented by the proposed model have significantly lower intensity errors, significantly higher peak signal-to-noise ratio, and significantly higher structural similarity with respect to the ground truth images. Besides, the proposed method significantly reduced the 3D distance of the CBCT-based tumor localization errors. In addition, the CBCT augmentation is nearly real-time. CONCLUSIONS: With the prior-image guidance, the proposed method is effective in reconstructing high-quality CBCT images from the highly under-sampled projections, considerably reducing the imaging dose and improving the clinical utility of the CBCT.

A generative adversarial network (GAN)-based technique for synthesizing realistic respiratory motion in the extended cardiac-torso (XCAT) phantoms.

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 thex/y/zdirection 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.

Radiogenomic Analysis of Locally Advanced Lung Cancer Based on CT Imaging and Intratreatment Changes in Cell-Free DNA.

The radiologic appearance of locally advanced lung cancer may be linked to molecular changes of the disease during treatment, but characteristics of this phenomenon are poorly understood. Radiomics, liquid biopsy of cell-free DNA (cfDNA), and next-generation sequencing of circulating tumor DNA (ctDNA) encode tumor-specific radiogenomic expression patterns that can be probed to study this problem. Preliminary findings are reported from a radiogenomic analysis of CT imaging, cfDNA, and ctDNA in 24 patients (median age, 64 years; range, 49-74 years) with stage III lung cancer undergoing chemoradiation on a prospective pilot study (NCT00921739) between September 2009 and September 2014. Unsupervised clustering of radiomic signatures resulted in two clusters that were associated with ctDNA TP53 mutations (P = .03) and changes in cfDNA concentration after 2 weeks of chemoradiation (P = .02). The radiomic features dissimilarity (hazard ratio [HR] = 0.56; P = .05), joint entropy (HR = 0.56; P = .04), sum entropy (HR = 0.53; P = .02), and normalized inverse difference (HR = 1.77; P = .05) were associated with overall survival. These results suggest heterogeneous and low-attenuating disease without a detectable ctDNA TP53 mutation was associated with early surges of cfDNA concentration in response to therapy and a generally better prognosis. Keywords: CT-Quantitative, Radiation Therapy, Lung, Computer Applications-3D, Oncology, Tumor Response, Outcomes Analysis Clinical trial registration no. NCT00921739 Supplemental material is available for this article. © RSNA, 2021.

Intrinsic radiomic expression patterns after 20 Gy demonstrate early metabolic response of oropharyngeal cancers.

PURPOSE: This study investigated the prognostic potential of intra-treatment PET radiomics data in patients undergoing definitive (chemo) radiation therapy for oropharyngeal cancer (OPC) on a prospective clinical trial. We hypothesized that the radiomic expression of OPC tumors after 20 Gy is associated with recurrence-free survival (RFS). MATERIALS AND METHODS: Sixty-four patients undergoing definitive (chemo)radiation for OPC were prospectively enrolled on an IRB-approved study. Investigational 18 F-FDG-PET/CT images were acquired prior to treatment and 2 weeks (20 Gy) into a seven-week course of therapy. Fifty-five quantitative radiomic features were extracted from the primary tumor as potential biomarkers of early metabolic response. An unsupervised data clustering algorithm was used to partition patients into clusters based only on their radiomic expression. Clustering results were naïvely compared to residual disease and/or subsequent recurrence and used to derive Kaplan-Meier estimators of RFS. To test whether radiomic expression provides prognostic value beyond conventional clinical features associated with head and neck cancer, multivariable Cox proportional hazards modeling was used to adjust radiomic clusters for T and N stage, HPV status, and change in tumor volume. RESULTS: While pre-treatment radiomics were not prognostic, intra-treatment radiomic expression was intrinsically associated with both residual/recurrent disease (P = 0.0256, χ 2 test) and RFS (HR = 7.53, 95% CI = 2.54-22.3; P = 0.0201). On univariate Cox analysis, radiomic cluster was associated with RFS (unadjusted HR = 2.70; 95% CI = 1.26-5.76; P = 0.0104) and maintained significance after adjustment for T, N staging, HPV status, and change in tumor volume after 20 Gy (adjusted HR = 2.69; 95% CI = 1.03-7.04; P = 0.0442). The particular radiomic characteristics associated with outcomes suggest that metabolic spatial heterogeneity after 20 Gy portends complete and durable therapeutic response. This finding is independent of baseline metabolic imaging characteristics and clinical features of head and neck cancer, thus providing prognostic advantages over existing approaches. CONCLUSIONS: Our data illustrate the prognostic value of intra-treatment metabolic image interrogation, which may potentially guide adaptive therapy strategies for OPC patients and serve as a blueprint for other disease sites. The quality of our study was strengthened by its prospective image acquisition protocol, homogenous patient cohort, relatively long patient follow-up times, and unsupervised clustering formalism that is less prone to hyper-parameter tuning and over-fitting compared to supervised learning.

Development of realistic multi-contrast textured XCAT (MT-XCAT) phantoms using a dual-discriminator conditional-generative adversarial network (D-CGAN).

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.

Dose-Distribution-Driven PET Image-Based Outcome Prediction (DDD-PIOP): A Deep Learning Study for Oropharyngeal Cancer IMRT Application.

PURPOSE: To develop a deep learning-based AI agent, DDD-PIOP (Dose-Distribution-Driven PET Image Outcome Prediction), for predicting 18FDG-PET image outcomes of oropharyngeal cancer (OPC) in response to intensity-modulated radiation therapy (IMRT). METHODS: DDD-PIOP uses pre-radiotherapy 18FDG-PET/CT images and the planned spatial dose distribution as the inputs, and it predicts the 18FDG-PET image outcomes in response to the planned IMRT delivery. This AI agent centralizes a customized convolutional neural network (CNN) as a deep learning approach, and it incorporates a few designs to enhance prediction accuracy. 66 OPC patients who received IMRT treatment on a sequential boost regime (2 Gy/daily fraction) were studied for DDD-PIOP development. 61 patients were used for AI agent training/validation, and the remaining five were used as independent tests. To evaluate the developed AI agent's performance, the predicted mean standardized uptake values (SUVs) of gross tumor volume (GTV) and clinical target volume (CTV) were compared with the ground truth values. Overall SUV distribution accuracy was evaluated by gamma test passing rates under different criteria. RESULTS: The developed DDD-PIOP successfully generated 18FDG-PET image outcome predictions for five test patients. The predicted mean SUV values of GTV/CTV were 3.50/1.41, which were close to the ground-truth values of 3.57/1.51. In 2D-based gamma tests, the average passing rate was 92.1% using 5%/10 mm criteria, which was improved to 95.9%/93.2% when focusing on GTV/CTV regions. 3D gamma test passing rates were 98.7% using 5%/10 mm criteria, and the corresponding GTV/CTV results were 99.8%/99.4%. CONCLUSION: The reported results suggest that the developed AI agent DDD-PIOP successfully predicted 18FDG-PET image outcomes with high quantitative accuracy. The generated voxel-based image outcome predictions could be used for treatment planning optimization prior to radiation delivery for the best individual-based outcome.

Automatic IMRT planning via static field fluence prediction (AIP-SFFP): a deep learning algorithm for real-time prostate treatment planning.

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 PTV58.8 Gy/PTV70 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 (PTV70 Gy V70 Gy = 95%), all 14 AI plans met institutional criteria. The coverage of PTV58.8 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.

Digital phantoms for characterizing inconsistencies among radiomics extraction toolboxes.

PURPOSE: to develop digital phantoms for characterizing inconsistencies among radiomics extraction methods based on three radiomics toolboxes: CERR (Computational Environment for Radiological Research), IBEX (imaging biomarker explorer), and an in-house radiomics platform. MATERIALS AND METHODS: we developed a series of digital bar phantoms for characterizing intensity and texture features and a series of heteromorphic sphere phantoms for characterizing shape features. The bar phantoms consisted of n equal-width bars (n = 2, 4, 8, or 64). The voxel values of the bars were evenly distributed between 1 and 64. Starting from a perfect sphere, the heteromorphic sphere phantoms were constructed by stochastically attaching smaller spheres to the phantom surface over 5500 iterations. We compared 61 features typically extracted from three radiomics toolboxes: (1) CERR (2) IBEX (3) in-house toolbox. The degree of inconsistency was quantified by concordance correlation coefficient (CCC) and Pearson correlation coefficient (PCC). Sources of discrepancies were characterized based on differences in mathematical definition, pre-processing, and calculation methods. RESULTS: For the intensity and texture features, only 53%, 45%, 55% features demonstrated perfect reproducibility (CCC = 1) between in-house/CERR, in-house/IBEX, and CERR/IBEX comparisons, while 71%, 61%, 61% features reached CCC > 0.8 and 25%, 39%, 39% features were with CCC < 0.5, respectively. Meanwhile, most features demonstrated PCC > 0.95. For shape features, the toolboxes produced similar (CCC > 0.98) volume yet inconsistent surface area, leading to inconsistencies in other shape features. However, all toolboxes resulted in PCC > 0.8 for all shape features except for compactness 1, where inconsistent mathematical definitions were observed. Discrepancies were characterized in pre-processing and calculation implementations from both type of phantoms. CONCLUSIONS: Inconsistencies among radiomics extraction toolboxes can be accurately identified using the developed digital phantoms. The inconsistencies demonstrate the significance of implementing quality assurance (QA) of radiomics extraction for reproducible and generalizable radiomic studies. Digital phantoms are therefore very useful tools for QA.

An investigation of machine learning methods in delta-radiomics feature analysis.

PURPOSE: This study aimed to investigate the effectiveness of using delta-radiomics to predict overall survival (OS) for patients with recurrent malignant gliomas treated by concurrent stereotactic radiosurgery and bevacizumab, and to investigate the effectiveness of machine learning methods for delta-radiomics feature selection and building classification models. METHODS: The pre-treatment, one-week post-treatment, and two-month post-treatment T1 and T2 fluid-attenuated inversion recovery (FLAIR) MRI were acquired. 61 radiomic features (intensity histogram-based, morphological, and texture features) were extracted from the gross tumor volume in each image. Delta-radiomics were calculated between the pre-treatment and post-treatment features. Univariate Cox regression and 3 multivariate machine learning methods (L1-regularized logistic regression [L1-LR], random forest [RF] or neural networks [NN]) were used to select a reduced number of features, and 7 machine learning methods (L1-LR, L2-LR, RF, NN, kernel support vector machine [KSVM], linear support vector machine [LSVM], or naïve bayes [NB]) was used to build classification models for predicting OS. The performances of the total 21 model combinations built based on single-time-point radiomics (pre-treatment, one-week post-treatment, and two-month post-treatment) and delta-radiomics were evaluated by the area under the receiver operating characteristic curve (AUC). RESULTS: For a small cohort of 12 patients, delta-radiomics resulted in significantly higher AUC than pre-treatment radiomics (p-value<0.01). One-week/two-month delta-features resulted in significantly higher AUC (p-value<0.01) than the one-week/two-month post-treatment features, respectively. 18/21 model combinations were with higher AUC from one-week delta-features than two-month delta-features. With one-week delta-features, RF feature selector + KSVM classifier and RF feature selector + NN classifier showed the highest AUC of 0.889. CONCLUSIONS: The results indicated that delta-features could potentially provide better treatment assessment than single-time-point features. The treatment assessment is substantially affected by the time point for computing the delta-features and the combination of machine learning methods for feature selection and classification.