Qifan Xu

While 4 π radiotherapy has shown dosimetric advantages over coplanar delivery, prior studies have assumed a fixed source-to-isocenter distance (SID), as varying SID is cumbersome to implement on conventional C-arm linacs. Radiotherapy platforms based on articulated robotic arms eliminate this constraint, enabling non-coplanar, non-isocentric delivery with variable SIDs without moving the patient. Larger SIDs offer increased per-field tumor coverage with a larger field-of-view (FOV), while shorter SIDs improve modulation resolution. This study combines multiple SIDs to simultaneously achieve high target coverage and modulation resolution in robotic radiotherapy. A Multi-SID optimization framework was developed for simultaneous beam orientation optimization, isocenter selection, SID selection, and fluence map optimization. The planning cost includes a least-squares dose fidelity objective, a total variation term for fluence smoothness, and a group sparsity term for beam orientation, SID, and isocenter selection. The collision-free candidate beam pool includes over 1000 beam orientations, 10 SID levels (50–95 cm), and 5–9 isocenters. Multi-SID 4 π intensity-modulated radiation therapy (IMRT) plans were compared against fixed-SID (SID-50 cm and SID-100 cm) 4 π IMRT plans on 10 head-and-neck (H&N) cancer cases. Multi-SID consistently achieved superior plan quality across all 10 patients using varying beams (10–30). For 15-beam plans, the planning target volume homogeneity (defined as D95/D5) is (0.898, 0.893, 0.889), dose conformity (using the Paddick Conformity Index) is (0.867, 0.853, 0.837), the mean Organs-at-Risk (OAR) dose is (12.67, 15.05, 14.12), and the R50 is (2.429, 2.575, 2.485) for Multi-SID, SID-100, and SID-50 respectively. Wilcoxon signed-rank testing showed that the improvements in homogeneity, conformity, and OAR sparing with Multi-SID were statistically significant. Our Multi-SID method leverages both high target coverage and modulation resolution to achieve better dosimetry and/or more efficient treatment plans than fixed SID methods.

Background. Preclinical small-animal studies are essential for understanding radiation-induced biological responses and toxicities, ultimately facilitating translation to clinical interventions. However, the promise is hampered by poor dose conformity achieved in small-animal experiments. The gap has been partially closed with recent advances in image guidance and radiation intensity modulation. The vast differences in the sizes between the small animals and humans, however, demand steeper dose gradient to mimic clinical targeting specificity. Objective. Here, we explore the non-coplanar solution space enabled by the Small Animal Radiation Research Platform (SARRP) to enhance dose delivery specificity. Approach. This study integrates a 4π non-coplanar beam orientation optimization framework with 4π rectangular apertures-based beam orientation optimization (4π-RABOO) for automated small-animal intensity-modulated radiotherapy (IMRT) planning. Methods. 4π-RABOO objective function included an L ₂-norm term for dose fidelity, an L ₁-norm term for aperture sparsity, and an L _2,1/2 group sparsity term for beam selection. Solving the large-scale computational problem was accelerated by replacing the Kronecker product with smaller matrix multiplications and by applying the Fast Iterative Shrinkage Thresholding Algorithm. Three preclinical sites (brain, liver tumor, spine) involving five mice were evaluated against two reference coplanar plans: 5-field rectangular aperture plans and fixed-size aperture plans. Plan evaluation metrics included D2, D98, homogeneity index, R ₅₀ for planning target volumes, and mean/max doses to organs at risk (OARs). Main Results. 4π-RABOO improved plan quality by reducing D2 by up to 17.4%, increasing D98 by up to 17.1%, and lowering R ₅₀ by over 30% compared to baseline plans. In spine cases, mean doses to critical OARs (bowel, liver, spleen) were reduced by up to 64.5%, 51.6%, and 93.1%, respectively. The enhanced dosimetric performance came at the cost of increased plan complexity, with 2–3X more apertures per case. Significance. Integrating beam orientation and rectangular aperture optimization into small-animal IMRT on the SARRP generates dosimetrically superior plans, significantly improving target dose homogeneity and adjacent tissue sparing.

Purpose. In radiotherapy, dose distribution conformity and compactness are critical to patient outcomes. Advanced techniques like 4 π radiotherapy leverage non-coplanar beams for superior dosimetry by exploring additional degrees of freedom. However, 4 π planning is computationally intensive due to large dose-loading matrices for candidate beams. This work presents an ultra-high performance parallel (UHPP) framework to accelerate high-dimensional treatment planning. Methods. For dose calculation, we developed: (1) a two-step total energy released per unit mass (TERMA) computation module calculating the TERMA array once per beam, enabling reuse across convolution directions; (2) a synchronized dose calculation module based on collapsed-cone convolution superposition (CCCS), arranging rays in dedicated sequences to preserve thread efficiency and minimize memory access; (3) a scattering-based coordinate transformation mapping dose from beamlet to patient Cartesian coordinates, eliminating aliasing without atomic operations. The framework includes CCCS exponential kernel calculation for varying LINAC spectra. For beam orientation optimization, we employed fast iterative shrinkage-thresholding algorithm with group sparsity regularization, accelerated using cuSPARSE library on GPUs. We benchmarked against Monte Carlo (MC) simulations for dose accuracy and compared computational performance to state-of-the-art (SOTA) methods. Plan quality was evaluated across four approaches: UHPP, SOTA, clinical VMAT plans, and MC calculations based on UHPP plans. Results. Compared to MC simulations, UHPP achieved minimum 98% gamma passing rates under 1.5%/1.5 mm criterion for water and slab phantoms, and average 97.35% and 92.18% under 3%/3 mm criterion for pancreas and head-and-neck patients, respectively. UHPP delivered 8.86× and 6.99× speedups in dose calculation and plan optimization while maintaining comparable or superior plan quality. Both UHPP and SOTA consistently produced 4 π plans outperforming clinical VMAT plans in organ-at-risk sparing and target coverage. Conclusion. The UHPP framework delivers high dose accuracy and substantial computational speedup without sacrificing 4 π planning’s dosimetric advantages, supporting practical adoption of advanced 4 π radiotherapy in clinical workflows.

Objective. Lung tumors can be obscured in x-rays, preventing accurate and robust localization. To improve lung conspicuity for image-guided procedures, we isolate the lungs in the anterior-posterior (AP) x-rays using a lung extraction network (LeX-net) that virtually removes overlapping thoracic structures, including ribs, diaphragm, liver, heart, and trachea. Approach. 73 965 thoracic 3DCTs and 106 thoracic 4DCTs were included. The 3D lung volume was extracted using an open-source lung volume segmentation model. AP digitally reconstructed radiographs (DRRs) of the full anatomy CT and extracted lungs were computed as the input and reference to train a network (LeX-net) to generate lung-extracted DRRs (LeX-net DRRs) from full anatomy DRRs, which adopted a Swin-UNet model with conditional GAN. Subsequently, the trained LeX-net on 3DCT was applied to 4DCT-derived DRRs. Lung tumor tracking was then performed on DRRs using a template-matching method on a holdoff 4DCT test set of 79 patients whose gross tumor volumes were smaller than 20 cm³. Main results. LeX-net successfully isolated the lungs in DRRs, achieving an SSIM of 0.9581 ± 0.0151 and a PSNR of 30.78 ± 2.50 on the testing set of 3DCT-derived DRRs. Its performance declined slightly when applied to the 4DCT but maintained useable lung-only 2D views. On the challenging test set including cases of organ overlap, high tumor mobility, and small tumor size, the individual tumor tracking error for LeX-net DRRs was 0.97 ± 0.86 mm, significantly lower than that of 3.13 ± 5.82 mm using the full anatomy DRRs. LeX-net improved success rates of using 5 mm, 3 mm, and 1 mm tracking windows from 88.1%, 80.0%, and 58.7% to 98.1%, 94.2%, and 73.8%, respectively. Significance. LeX-net removes overlapping anatomies and enhances visualization of the lungs in x-rays. The model trained using 3DCTs is generalizable to 4DCT-derived DRRs, achieving significantly improved tumor tracking outcome.

Image-guided mouse irradiation is essential to understand interventions involving radiation prior to human studies. Our objective is to employ Swin UNEt TRansformers (Swin UNETR) to segment native micro-CT and contrast-enhanced micro-CT scans and benchmark the results against 3D no-new-Net (nnU-Net). Swin UNETR reformulates mouse organ segmentation as a sequence-to-sequence prediction task using a hierarchical Swin Transformer encoder to extract features at five resolution levels, and it connects to a Fully Convolutional Neural Network (FCNN)-based decoder via skip connections. The models were trained and evaluated on open datasets, with data separation based on individual mice. Further evaluation on an external mouse dataset acquired on a different micro-CT with lower kVp and higher imaging noise was also employed to assess model robustness and generalizability. The results indicate that Swin UNETR consistently outperforms nnU-Net and AIMOS in terms of the average dice similarity coefficient (DSC) and the Hausdorff distance (HD95p), except in two mice for intestine contouring. This superior performance is especially evident in the external dataset, confirming the model’s robustness to variations in imaging conditions, including noise and quality, and thereby positioning Swin UNETR as a highly generalizable and efficient tool for automated contouring in pre-clinical workflows.

Suppression of thoracic bone shadows on chest X-rays (CXRs) has been indicated to improve the diagnosis of pulmonary disease. Previous approaches can be categorized as unsupervised physical and supervised deep learning models. Nevertheless, with physical models able to preserve morphological details but at the cost of extremely long processing time, existing DL methods face challenges of gathering sufficient/qualitative ground truth (GT) for robust training, thus leading to failure in maintaining clinically acceptable false positive rates. We hereby propose a generalizable yet efficient workflow of two stages: (1) training pairs generation with GT bone shadows eliminated in by a physical model in spatially transformed gradient fields. (2) fully supervised image denoising network training on stage-one datasets for fast rib removal on incoming CXRs. For step two, we designed a densely connected network called SADXNet, combined with peak signal to noise ratio and multi-scale structure similarity index measure objective minimization to suppress bony structures. The SADXNet organizes spatial filters in U shape (e.g., X=7; filters = 16, 64, 256, 512, 256, 64, 16) and preserves the feature map dimension throughout the network flow. Visually, SADXNet can suppress the rib edge and that near the lung wall/vertebra without jeopardizing the vessel/abnormality conspicuity. Quantitively, it achieves RMSE of ~0 during testing with one prediction taking <1s. Downstream tasks including lung nodule detection as well as common lung disease classification and localization are used to evaluate our proposed rib suppression mechanism. We observed 3.23% and 6.62% area under the curve (AUC) increase as well as 203 and 385 absolute false positive decrease for lung nodule detection and common lung disease localization, separately.

The Plug-and-Play (PnP) framework was recently introduced for low-dose CT reconstruction to leverage the interpretability and the flexibility of model-based methods to incorporate various plugins, such as trained deep learning (DL) neural networks. However, the benefits of PnP vs. state-of-the-art DL methods have not been clearly demonstrated. In this work, we proposed an improved PnP framework to address the previous limitations and develop clinical-relevant segmentation metrics for quantitative result assessment. Compared with the DL alone methods, our proposed PnP framework was slightly inferior in MSE and PSNR. However, the power spectrum of the resulting images better matched that of full-dose images than that of DL denoised images. The resulting images supported higher accuracy in airway segmentation than DL denoised images for all the ten patients in the test set, more substantially on the airways with a cross-section smaller than 0.61cm ² , and outperformed the DL denoised images for 45 out of 50 lung lobes in lobar segmentation. Our PnP method proved to be significantly better at preserving the image texture, which translated to task-specific benefits in automated structure segmentation and detection.