Hypoxia Informed RBE-Weighted Beam Orientation Optimization for Intensity Modulated Proton Therapy Using [18F]-FMISO-PET Estimation of pO2

PURPOSE/OBJECTIVE(S)

Variable relative biological effectiveness (RBE) models have previously informed proton therapy dose optimization algorithms, but few models have incorporated hypoxia's increase on radioresistance. Here, we obtain voxel-based estimation of partial oxygen pressure to weigh RBE values in a single biologically informed beam orientation optimization (BOO) algorithm.

MATERIALS/METHODS

Four brain cancer patients with [18F]-FMISO-PET/CT images were selected from an HCP database. Oxygen values were derived from tracer uptake using a non-linear least squares curve fitting. RBE dose was then weighted using oxygen enhancement ratios (OER) for each structure and substituted into the dose fidelity term of our BOO algorithm. The nonlinear optimization problem was solved using a split-Bregman approach, with FISTA as the solver. This method (HypRBE) was compared dose fidelity terms using the Rorvik RBE model (RegRBE), without OER. Tumor homogeneity index (HI), Dmax, and D95% were evaluated along with worst-case statistics after normalization to normal tissue isotoxicity.

RESULTS

Compared to RegRBE, HypRBE increased tumor [HI, Dmax, D95%] on average by [0.5%, 2.0%, 2.5%] and improved worst-case tumor [HI, Dmax, D95%] by [5.3%, 16.2%, 9.6%]. HypRBE shows an increase in therapeutic ratio, and is notably robust against uncertainty scenarios.

CONCLUSION

We have developed an optimization algorithm whose dose fidelity term is weighted by hypoxia informed RBE values. We have shown that HypRBE selects beams that are better suited to protect low RBE, well-oxygenated normal tissue while maintaining high dose to high RBE, hypoxic tumor cells.

Contouring Analysis on Synthetic Contrast-Enhanced MR from GRMM-GAN and Implications on MR-Guide Radiation Therapy

PURPOSE/OBJECTIVE(S)

MR-guided linear accelerators have been commercialized making MR-only planning and adaptation an appealing alternative circumventing MR-CT registration. However, obtaining daily contrast-enhanced MR images can be prohibitive due to the increased risk of side effects from repeated contrast injections. In this work, we evaluate the quality of contrast-enhanced multi-modal MR image synthesis network GRMM-GAN (gradient regularized multi-modal multi-discrimination sparse-attention fusion generative adversarial network) for MR-guided radiation therapy.

MATERIALS/METHODS

With IRB approval, we trained the GRMM-GAN based on 165 abdominal MR studies from 65 patients. Each study included T2, T1 pre-contrast (T1pre), and T1 contrast enhanced (T1ce) images. The two pre-contrast MR modalities, T2 and T1pre images were adopted as inputs for GRMM-GAN, and the T1ce image at the portal venous phase was used as an output. Ten MR scans containing 21 liver tumors were selected for contouring analysis. A Turing test was first given to six radiation oncologists, in which 100 real T1ce and synthetic T1ce image slices are randomly given to the radiation oncologists to determine the authenticity of the synthesis. We then invited two radiation oncologists (RadOnc 1 and RadOnc2) to manually contour the 21 liver tumors independently on the real T1ce images. RadOnc2 then performed contouring on the respective synthetic T1ce MRs. DICE coefficient (defined as the intersection over the average of two volumes) and Hausdorff distance (HD, measuring how far two volumes are from each other) were used as analysis metrics. The DICE coefficients were calculated from the two radiation oncologists' contours on the real T1ce MR for each tumor. The DICE coefficients were also calculated from RadOnc 2's contours on real and synthetic MRs. Besides, tumor center shifts were extracted. The tumor center of mass coordinates was extracted from real and synthetic volumes. The difference in the coordinates indicated the shifts in the superior-inferior (SI), right-left (RL), and anterior-posterior (AP) directions between real and synthetic tumor volumes.

RESULTS

An average of 52.3% test score was achieved from the six radiation oncologists, which is close to random guessing. RadOnc 1 and RadOnc 2, who had participated in the contouring analysis, achieved an average DICE of 0.91±0.02 from tumor volumes drawn on the real T1ce MRs. This result sets the inter-operator uncertainty baseline in the real clinical setting. RadOnc 2 achieved an average DICE (real vs. synth) of 0.90±0.04 and HD of 4.76±1.82 mm. Only sub-millimeter (SI: 0.67 mm, RL: 0.41 mm, AP: 0.39 mm) tumor center shifts were observed in all three directions.

CONCLUSION

The GRMM-GAN method has the potential for MR-guided liver radiation when contrast agents cannot be administered daily and provide synthetic contrast-enhanced MR for better tumor targeting. The network can produce synthetic MR images with satisfactory contour agreement and geometric integrity.

Towards Automated Dosimetric Analysis of the Bladder Trigone: Deep-Learning-Based Joint Segmentation and Landmark Localization

PURPOSE/OBJECTIVE(S)

The bladder trigone dosimetry is hypothesized to have a stronger correlation with post-SBRT urinary toxicity than that of the entire bladder. However, the trigone tends to move significantly between simulation and daily treatment. Its small size, large daily motion, and proximity to the target lead to potentially consequential but unaccounted-for dosimetric uncertainties. Manual segmentation of the structure can be inconsistent and time-consuming, even with MR-guided RT. Here, we propose and demonstrate a deep-learning-based framework for joint segmentation and landmark localization to support deformable registration and comprehensive dosimetric analysis.

MATERIALS/METHODS

A total of 30 patients were randomly selected for training, and 20 were held out for testing. Each patient had 1 simulation and 5 daily pre-treatment images obtained from a clinical 0.35T MR Linac. The trigone is defined as the triangular bladder section among three landmarks (2 ureteral orifices and the internal urethral orifice). In the manual contouring process, the 3 landmarks were identified first, followed by trigone segmentation. The proposed joint method uses a modified 3D nnU-Net with 2 decoders, one for segmentation and the other for landmark localization. The shared encoder is expected to extract features useful for both tasks. The joint framework was compared with a baseline method using two separate 3D nnU-Nets for landmark localization and trigone segmentation, respectively. Since the trigone is small (∼2% of the bladder volume), we further experimented with a second-stage prediction mimicking the human contouring process. The predicted landmarks from the first stage were used to crop the trigone region, and a second network was trained on cropped images. Evaluation metrics included the Dice score, 95% Hausdorff distance (HD95), and average surface distance (ASD) for segmentation, and Euclidean distance (ED) between the predicted and ground truth landmarks for localization.

RESULTS

The quantification metrics are summarized in the table below. The joint approach shows similar Dice performance to the baseline method but markedly better HD95 by 13%. For landmark localization, the proposed method is similar to the baseline, but the integration of the segmentation task stabilizes the training process. The two-stage approach further improves HD95, ASD, and ED by 27%, 24%, and 19%.

CONCLUSION

Combining segmentation and landmark localization exhibits a synergistic effect. The proposed two-stage approach provided additional improvement. Future studies will explore the deformable registration of the trigone based on the segmentation and landmark detection, as well as analyze cumulated dose distribution.

Artificial Intelligence for Automated DWI/FLAIR Mismatch Assessment on Magnetic Resonance Imaging in Stroke: A Systematic Review

We conducted this Systematic Review to create an overview of the currently existing Artificial Intelligence (AI) methods for Magnetic Resonance Diffusion-Weighted Imaging (DWI)/Fluid-Attenuated Inversion Recovery (FLAIR)-mismatch assessment and to determine how well DWI/FLAIR mismatch algorithms perform compared to domain experts. We searched PubMed Medline, Ovid Embase, Scopus, Web of Science, Cochrane, and IEEE Xplore literature databases for relevant studies published between 1 January 2017 and 20 November 2022, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We assessed the included studies using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. Five studies fit the scope of this review. The area under the curve ranged from 0.74 to 0.90. The sensitivity and specificity ranged from 0.70 to 0.85 and 0.74 to 0.84, respectively. Negative predictive value, positive predictive value, and accuracy ranged from 0.55 to 0.82, 0.74 to 0.91, and 0.73 to 0.83, respectively. In a binary classification of ±4.5 h from stroke onset, the surveyed AI methods performed equivalent to or even better than domain experts. However, using the relation between time since stroke onset (TSS) and increasing visibility of FLAIR hyperintensity lesions is not recommended for the determination of TSS within the first 4.5 h. An AI algorithm on DWI/FLAIR mismatch assessment focused on treatment eligibility, outcome prediction, and consideration of patient-specific data could potentially increase the proportion of stroke patients with unknown onset who could be treated with thrombolysis.

Automated Identification of Multiple Findings on Brain MRI for Improving Scan Acquisition and Interpretation Workflows: A Systematic Review

We conducted a systematic review of the current status of machine learning (ML) algorithms’ ability to identify multiple brain diseases, and we evaluated their applicability for improving existing scan acquisition and interpretation workflows. PubMed Medline, Ovid Embase, Scopus, Web of Science, and IEEE Xplore literature databases were searched for relevant studies published between January 2017 and February 2022. The quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. The applicability of ML algorithms for successful workflow improvement was qualitatively assessed based on the satisfaction of three clinical requirements. A total of 19 studies were included for qualitative synthesis. The included studies performed classification tasks (n = 12) and segmentation tasks (n = 7). For classification algorithms, the area under the receiver operating characteristic curve (AUC) ranged from 0.765 to 0.997, while accuracy, sensitivity, and specificity ranged from 80% to 100%, 72% to 100%, and 65% to 100%, respectively. For segmentation algorithms, the Dice coefficient ranged from 0.300 to 0.912. No studies satisfied all clinical requirements for successful workflow improvements due to key limitations pertaining to the study’s design, study data, reference standards, and performance reporting. Standardized reporting guidelines tailored for ML in radiology, prospective study designs, and multi-site testing could help alleviate this.

A nonvoxel-based dose convolution/superposition algorithm optimized for scalable GPU architectures

PURPOSE

Real-time adaptive planning and treatment has been infeasible due in part to its high computational complexity. There have been many recent efforts to utilize graphics processing units (GPUs) to accelerate the computational performance and dose accuracy in radiation therapy. Data structure and memory access patterns are the key GPU factors that determine the computational performance and accuracy. In this paper, the authors present a nonvoxel-based (NVB) approach to maximize computational and memory access efficiency and throughput on the GPU.

METHODS

The proposed algorithm employs a ray-tracing mechanism to restructure the 3D data sets computed from the CT anatomy into a nonvoxel-based framework. In a process that takes only a few milliseconds of computing time, the algorithm restructured the data sets by ray-tracing through precalculated CT volumes to realign the coordinate system along the convolution direction, as defined by zenithal and azimuthal angles. During the ray-tracing step, the data were resampled according to radial sampling and parallel ray-spacing parameters making the algorithm independent of the original CT resolution. The nonvoxel-based algorithm presented in this paper also demonstrated a trade-off in computational performance and dose accuracy for different coordinate system configurations. In order to find the best balance between the computed speedup and the accuracy, the authors employed an exhaustive parameter search on all sampling parameters that defined the coordinate system configuration: zenithal, azimuthal, and radial sampling of the convolution algorithm, as well as the parallel ray spacing during ray tracing. The angular sampling parameters were varied between 4 and 48 discrete angles, while both radial sampling and parallel ray spacing were varied from 0.5 to 10 mm. The gamma distribution analysis method (γ) was used to compare the dose distributions using 2% and 2 mm dose difference and distance-to-agreement criteria, respectively. Accuracy was investigated using three distinct phantoms with varied geometries and heterogeneities and on a series of 14 segmented lung CT data sets. Performance gains were calculated using three 256 mm cube homogenous water phantoms, with isotropic voxel dimensions of 1, 2, and 4 mm.

RESULTS

The nonvoxel-based GPU algorithm was independent of the data size and provided significant computational gains over the CPU algorithm for large CT data sizes. The parameter search analysis also showed that the ray combination of 8 zenithal and 8 azimuthal angles along with 1 mm radial sampling and 2 mm parallel ray spacing maintained dose accuracy with greater than 99% of voxels passing the γ test. Combining the acceleration obtained from GPU parallelization with the sampling optimization, the authors achieved a total performance improvement factor of >175 000 when compared to our voxel-based ground truth CPU benchmark and a factor of 20 compared with a voxel-based GPU dose convolution method.

CONCLUSIONS

The nonvoxel-based convolution method yielded substantial performance improvements over a generic GPU implementation, while maintaining accuracy as compared to a CPU computed ground truth dose distribution. Such an algorithm can be a key contribution toward developing tools for adaptive radiation therapy systems.