Potential patient-care applications include radiotherapy planning, clinical studies of neurodegenerative diseases and real-time image-guided navigation for surgery.

Jian Wang has developed a deep learning model for predictive 3D image registration, with potential to improve the efficiency of large-scale image analysis for radiotherapy planning, clinical studies of neurodegenerative diseases and real-time image-guided navigation for surgery. Wang, a Ph.D. student of Computer Science at the University of Virginia, developed the model with his advisor Miaomiao Zhang, assistant professor of Electrical and Computer Engineering and Computer Science.

Wang and Zhang presented their co-authored paper, DeepFLASH: An Efficient Network for Learning-based Medical Image Registration, at the 2020 IEEE Computer Vision and Pattern Recognition Conference.

“I was honored to attend one of the most prestigious conferences in computer vision.” Wang said. “Brilliant talks presented by research scientists broadened my horizons on image analysis models and made me realize that some great ideas proposed by scientists are coming true. The CVPR experience enhanced my intent to be a diligent Ph.D. student with a modest research attitude.”

Computational benefits of Wang and Zhang’s model include greater speed and lower memory consumption for both model training and inference making, compared to state-of-the-art image registration techniques. Monitoring brain tumor removal surgery via magnetic resonance images or ultrasound scans, leading to improved brain surgery survival rates with lower risks of collateral tissue damage, is an ongoing research with patient-care application in Zhang’s group.