Radiation Medicine Commons^™

Blinded, Bias-Controlled Multi-Rater Evaluation Of Human-Versus-Ai Brain Metastasis Segmentation Using A Hybrid Foundation-Model Framework, Yiding Han, Enze Zhu, Mhd Hasan Ai Mekdash, Omar Awad, Piyush Pathak, Shixiao Liang, Daniel Allan Hamstra, Xizhe Zhang, Zaid Ali Siddiqui, Baozhou Sun

Faculty, Staff and Students Publications

Background: Accurate segmentation of brain metastases (BM) is essential for diagnosis, stereotactic radiosurgery planning, and longitudinal assessment. However, manual contouring is time-intensive, limiting clinical scalability, and exhibits substantial inter-observer variability. This variability complicates objective assessment of automated segmentation methods and challenges interpretation of model performance.

Purpose: To address these limitations, we developed TUM-SAM, a hybrid foundation-model framework for fully automated BM segmentation, and introduced a bias-controlled, blinded multi-rater evaluation paradigm to determine whether AI-based BM segmentation has reached expert-level performance and whether AI-generated contours are preferred by human experts under unbiased assessment.

Methods: TUM-SAM integrates nnU-Net-based lesion detection with a …

A Multi-Institutional Epid-Based 3d Dose Reconstruction Model In A Virtual Phantom For Standardized Patient Qa, Benchmarking And Auditing For Stereotactic Radiosurgery, Benjamin Zwan, Emily Searle, Cameron Stanton, Ching-Ling Teng, Seng Boh Lim, Andrew Dipuglia, Richard Popple, Michael Lovelock, Ashley Cullen, Conor K Mcgarry, Victoria Robinson, Sergei Zavgorodni, Baozhou Sun, Xiaodong Zhao, Matthew Schmidt, Juan-Francisco Calvo-Ortega, Gemma Warner, Andrew Cousins, Michael Barnes, Peter Greer

Faculty, Staff and Students Publications

Background and purpose: For single-isocentre multi-target (SIMT) stereotactic radiosurgery (SRS), benchmarking, auditing and inter-institutional standardisation of dose verification remain challenging as they require specialized equipment and expertise. This work aims to develop and validate an electronic portal imaging device (EPID)-based technique to determine 3D dose in a virtual spherical phantom which is applicable across institutions for SIMT SRS dose verification.

Materials and methods: Small-field output factors from 11 international centres were measured in-water for the TrueBeam linear accelerator, including jaw-defined and high-definition multi-leaf collimator (MLC) fields from 0.5 × 0.5 to 20 × 20 cm2 and depths from 1.5-20 cm. …

Predicting Ventilation From Single Breathing Phase Non-Contrast Ct Using Swin Transformers, Yi-Kuan Liu, Hsu-Ting Kuo, Alaa Melek, Richard Castillo, Yevgeniy Vinogradskiy, Lili Zhao, Girish Nair, Edward Castillo

Department of Radiation Oncology Faculty Papers

BACKGROUND: Pulmonary ventilation imaging enables functional avoidance radiotherapy treatment plans by quantifying regional lung function. However, current clinical standards, such as 99𝑚Tc-based single-photon emission computed tomography (SPECT), rely on radioactive tracers, which can introduce imaging deposition artifacts. CT ventilation imaging (CTVI) methods based on both physical models and deep learning approaches currently require multiple CT images as input, such as the inhale/exhale phases of a 4DCT. While the theoretical foundation of physics-based CTVI is built on multi-phase information, the feasibility of single-phase deep learning CTV models has not been determined.

PURPOSE: While deep learning methods have predicted SPECT ventilation from …

Computed Tomography For Four-Dimensional Dose Calculation: Effect Of Detector Array Length, Inhwan Yeo, Qianyi Xu

Department of Radiation Oncology Faculty Papers

PURPOSE: Four-dimensional computed tomography (4DCT) is susceptible to a geometrical error when respiration changes upon patient shift. 4DCT using 16-cm detector arrays, not needing the shift, has been shown to improve geometrical accuracy, compared with a conventional 4DCT (4-cm array), albeit with some Hounsfield unit discrepancies. The 4DCTs were validated for 4D planning.

MATERIALS AND METHODS: A lung-mimicking phantom containing a spherical target with ten respiratory traces and three tumor-shaped targets with their traces was respectively imaged when, to each of the ten-positions/phases, the lung was moved (1) step-wisely by helical scan at each movement (ground truth; 4DCT_GT), …

Detection Of Phase-Binning And Interpolation Artifacts In 4-Dimensional Computed Tomography Imaging Using Deep Learning And Rule-Based Approaches, Jorge Cisneros, Nathan H. Feldt, Yevgeniy Vinogradskiy, Richard Castillo, Edward Castillo

Department of Radiation Oncology Faculty Papers

BACKGROUND: Four-dimensional computed tomography (4DCT) imaging is a crucial component to lung cancer radiotherapy planning and enables CT-ventilation-based functional avoidance planning to mitigate radiation toxicity. However, 4DCT scans are frequently impaired by acquisition artifacts that corrupt downstream analyses that depend on lung segmentation and deformable image registration, such as CT-ventilation and dose accumulation.

PURPOSE: This study develops 3D deep learning models to identify phase-binning artifacts at the voxel level and a heuristic, rule-based method to identify interpolation slices within 4DCT images.

METHODS: We introduce a generator that systematically inserts synthetic phase-binning and interpolation artifacts into any artifact-free breathing phase obtained …

Evaluation Of Image Quality In Mobile Cone-Beam Ct With Dose Modulation Using Automatic Exposure Control: A Phantom Study, Keita Okazaki, Wenchao Cao, Reza Taleei, Firas Mourtada, Jun Li, Karen Mooney, Pramila Rani Anne, Yingxuan Chen

Department of Radiation Oncology Faculty Papers

BACKGROUND: The integration of mobile cone-beam computed tomography (CBCT) into brachytherapy workflows offers clinical advantages such as immediate verification of applicator placement and adaptive treatment planning. These benefits require sufficient image quality to delineate applicators, target volumes, and organs at risk. A systematic evaluation of automatic exposure control (AEC) settings, radiation dose, and image quality is essential to ensure clinically acceptable imaging while minimizing patient exposure.

PURPOSE: This study evaluates the characteristics of AEC and its impact on image quality and radiation dose in a mobile CBCT system used for brachytherapy.

METHODS: The Elekta ImagingRing CBCT system was used to …

Evaluating Artifact-Free Four-Dimensional Computer Tomography With 16 Cm Detector Array, Inhwan Yeo, Wei Nie, Jiajin Fan, Mindy Joo, Michael Correa, Qianyi Xu

Department of Radiation Oncology Faculty Papers

PURPOSE: To evaluate a 16 cm-array axial four-dimensional computer tomography (4DCT) in comparison with a 4 cm-array 4DCT in the presence of respiration irregularity.

METHOD: Ten traces of lung tumor motion from CyberKnife treatments were imported to move the lung cylinder, containing a spherical target, of a phantom. Images were acquired for the lung that moved to each of the 10-positions/phases (1) step-wisely by nominal helical scan at each movement (ground truth), (2) continuously by 4D scan with the 16 cm array, and (3) the same with the 4 cm array, involving table shift. Irregularities, consisting of baseline shift and/or …

Deep Learning-Based Auto-Segmentation For Liver Yttrium-90 Selective Internal Radiation Therapy, Jun Li, Wookjin Choi, Rani Anne

Department of Radiation Oncology Faculty Papers

The aim was to evaluate a deep learning-based auto-segmentation method for liver delineation in Y-90 selective internal radiation therapy (SIRT). A deep learning (DL)-based liver segmentation model using the U-Net3D architecture was built. Auto-segmentation of the liver was tested in CT images of SIRT patients. DL auto-segmented liver contours were evaluated against physician manually-delineated contours. Dice similarity coefficient (DSC) and mean distance to agreement (MDA) were calculated. The DL-model-generated contours were compared with the contours generated using an Atlas-based method. Ratio of volume (RV, the ratio of DL-model auto-segmented liver volume to manually-delineated liver volume), and ratio of activity (RA, …

Fully Automated Deep Learning Based Auto-Contouring Of Liver Segments And Spleen On Contrast-Enhanced Ct Images, Aashish C Gupta, Guillaume Cazoulat, Mais Al Taie, Sireesha Yedururi, Bastien Rigaud, Austin Castelo, John Wood, Cenji Yu, Caleb O'Connor, Usama Salem, Jessica Albuquerque Marques Silva, Aaron Kyle Jones, Molly Mcculloch, Bruno C Odisio, Eugene J Koay, Kristy K Brock

Faculty, Staff and Student Publications

Manual delineation of liver segments on computed tomography (CT) images for primary/secondary liver cancer (LC) patients is time-intensive and prone to inter/intra-observer variability. Therefore, we developed a deep-learning-based model to auto-contour liver segments and spleen on contrast-enhanced CT (CECT) images. We trained two models using 3d patch-based attention U-Net ([Formula: see text] and 3d full resolution of nnU-Net ([Formula: see text] to determine the best architecture ([Formula: see text]. BA was used with vessels ([Formula: see text] and spleen ([Formula: see text] to assess the impact on segment contouring. Models were trained, validated, and tested on 160 ([Formula: see text]), …

Quantitative Longitudinal Mapping Of Radiation-Treated Prostate Cancer Using Mr Fingerprinting With Radial Acquisition And Subspace Reconstruction, Victoria Y Yu, Ricardo Otazo, Can Wu, Ergys Subashi, Manuel Baumann, Peter Koken, Mariya Doneva, Peter Mazurkewitz, Daniel Shasha, Michael Zelefsky, Laura Cervino, Ouri Cohen

Faculty, Staff and Student Publications

MR fingerprinting (MRF) enables fast multiparametric quantitative imaging with a single acquisition and has been shown to improve diagnosis of prostate cancer. However, most prostate MRF studies were performed with spiral acquisitions that are sensitive to B0 inhomogeneities and consequent blurring. In this work, a radial MRF acquisition with a novel subspace reconstruction technique was developed to enable fast T1/T2 mapping in the prostate in under 4 min. The subspace reconstruction exploits the extensive temporal correlations in the MRF dictionary to pre-compute a low dimensional space for the solution and thus reduce the number of radial spokes to accelerate the …

Comparison Of Reconstructed Prompt Gamma Emissions Using Maximum Likelihood Estimation And Origin Ensemble Algorithms For A Compton Camera System Tailored To Proton Range Monitoring, Ingrid Valencia Lozano, George Dedes, Steve Peterson, Dennis Mackin, Andreas Zoglauer, Sam Beddar, Stephen Avery, Jerimy Polf, Katia Parodi

Faculty, Staff and Student Publications

Compton-based prompt gamma (PG) imaging is being investigated by several groups as a potential solution for in vivo range monitoring in proton therapy. The performance of this technique depends on the detector system as well as the ability of the reconstruction method to obtain good spatial resolution to establish a quantitative correlation between the PG emission and the proton beam range in the patient. To evaluate the feasibility of PG imaging for range monitoring, we quantitatively evaluated the emission distributions reconstructed by a Maximum Likelihood Expectation Maximization (MLEM) and a Stochastic Origin Ensemble (SOE) algorithm. To this end, we exploit …

Applicability And Usage Of Dose Mapping/Accumulation In Radiotherapy, Martina Murr, Kristy K Brock, Marco Fusella, Nicholas Hardcastle, Mohammad Hussein, Michael G Jameson, Isak Wahlstedt, Johnson Yuen, Jamie R Mcclelland, Eliana Vasquez Osorio

Faculty, Staff and Student Publications

Dose mapping/accumulation (DMA) is a topic in radiotherapy (RT) for years, but has not yet found its widespread way into clinical RT routine. During the ESTRO Physics workshop 2021 on "commissioning and quality assurance of deformable image registration (DIR) for current and future RT applications", we built a working group on DMA from which we present the results of our discussions in this article. Our aim in this manuscript is to shed light on the current situation of DMA in RT and to highlight the issues that hinder consciously integrating it into clinical RT routine. As a first outcome of …

An Untrained Deep Learning Method For Reconstructing Dynamic Mr Images From Accelerated Model-Based Data, Kalina P Slavkova, Julie C Dicarlo, Viraj Wadhwa, Sidharth Kumar, Chengyue Wu, John Virostko, Thomas E Yankeelov, Jonathan I Tamir

Faculty, Staff and Student Publications

PURPOSE: To implement physics-based regularization as a stopping condition in tuning an untrained deep neural network for reconstructing MR images from accelerated data.

METHODS: The ConvDecoder (CD) neural network was trained with a physics-based regularization term incorporating the spoiled gradient echo equation that describes variable-flip angle data. Fully-sampled variable-flip angle k-space data were retrospectively accelerated by factors of R = {8, 12, 18, 36} and reconstructed with CD, CD with the proposed regularization (CD + r), locally low-rank (LR) reconstruction, and compressed sensing with L1-wavelet regularization (L1). Final images from CD + r training were evaluated at the "argmin" of …

Multi-Organ Segmentation Of Abdominal Structures From Non-Contrast And Contrast Enhanced Ct Images, Cenji Yu, Chidinma P Anakwenze, Yao Zhao, Rachael M Martin, Ethan B Ludmir, Joshua S Niedzielski, Asad Qureshi, Prajnan Das, Emma B Holliday, Ann C Raldow, Callistus M Nguyen, Raymond P Mumme, Tucker J Netherton, Dong Joo Rhee, Skylar S Gay, Jinzhong Yang, Laurence E Court, Carlos E Cardenas

Faculty, Staff and Student Publications

Manually delineating upper abdominal organs at risk (OARs) is a time-consuming task. To develop a deep-learning-based tool for accurate and robust auto-segmentation of these OARs, forty pancreatic cancer patients with contrast-enhanced breath-hold computed tomographic (CT) images were selected. We trained a three-dimensional (3D) U-Net ensemble that automatically segments all organ contours concurrently with the self-configuring nnU-Net framework. Our tool's performance was assessed on a held-out test set of 30 patients quantitatively. Five radiation oncologists from three different institutions assessed the performance of the tool using a 5-point Likert scale on an additional 75 randomly selected test patients. The mean (± …