Medical Image Processing

Most medical image analysis applications require pre-processing steps including image segmentation (right) and deformable image registration (left). Registration is of particular importance the DMIC’s work on inferring patient-specific biomechanical properties from dynamic imaging.

In addition to biomechanical modeling, the development of medical image processing methods requires many aspects of applied mathematics, including the numerical solution of partial differential equations, iterative methods for large-scale linear systems, large-scale optimization, and parallel computing. Our recent work in this area has focused on how to incorporate deep learning methods into known algorithmic frameworks with well-defined numerical properties.

Selected Publications on Image Processing

Joshua Joseph, Aaron Luong, Debarghya Chaki, Jasmine Jung, Yi-Kuan Liu, Amanda Nowacki, Blake Evans, Hairong Wang, and Edward Castillo. Computationally Efficient Decoupled Momentum Optimization Algorithm for Medical Imaging Models. Scientific Reports. In Press, 2026.

Gabriela Roque Oliveira Nomura, Aaron Luong, Ananya Prakash, Annabelle Alemand, Tanish Bhowmick, Alisa Ali, Jaimie Ren, Basil Rehani, Girish Nair, Richard Castillo, Yevgeniy Vinogradskiy, and Edward Castillo. TriSwinUNETR Lobe Segmentation Model for Computing DIR-Free CT-Ventilation. Frontiers in Oncology, 15: 1475133, 2025.

Jorge Cisneros, Nathan Feldt, Yevgeniy Vinogradskiy, Richard Castillo, and Edward Castillo. Detection of Phase-Binning and Interpolation Artifacts in 4-Dimensional Computed Tomography Imaging Using Deep Learning and Rule-Based Approaches. Medical Physics. 52:e70191, 2025.

Charles Vu, Zaid Siddiqui, Leonid Zamdborg, Andrew Thompson, Thomas Quinn, Edward Castillo, and Thomas Guerrero. Deep Convolutional Neural Networks for Automatic Segmentation of Thoracic Organs-At-Risk in Radiation Oncology – Use of Non-Domain Transfer Learning. Journal of Applied Clinical Medical Physics, 21(6): 108-113, 2020.

Edward Castillo. Quadratic Penalty Method for Intensity-Based Deformable Image Registration and 4DCT Lung Motion Recovery. Medical Physics, 46(5): 2194-2203, 2019.