The University of Texas MD Anderson Cancer Center at Houston
Department of Radiation Physics

We work at the intersection of computational methods, medical imaging, and adaptive radiotherapy to make treatment safer, faster, more accessible, and more consistent. Modern imaging and computational techniques enable radiotherapy that is not only personalized but also responsive to day-to-day (even real-time) patient change. This creates challenges: more data, greater variability, tighter resources, and more decisions. Our group develops advanced computational methods, such as AI, to transform these data streams into radiation treatment across the pipeline: patient modeling, diagnosis; image acquisition/reconstruction; contouring; planning/replanning; and outcome/toxicity prediction.

A second focus is medical AI safety. Medical AI is a system, not just a model; failures with clinical impact often emerge at the interfaces among algorithms, users, and workflows, with risks amplified by nowadays generative models (e.g., LLMs, synthetic imaging, etc). We aim to optimize the workflow and design prospective quality-assurance layers that monitor and verify AI outputs to protect patients against such harm.

In parallel, we are also deeply engaged in implementing adaptive radiotherapy, CBCT-guided linac, MR-guided linac, and other modalities. Our goal is to pair robust computational methods with safety layer to accelerate, stabilize, and democratize adaptive radiotherapy across clinics while maintaining rigorous safety and consistency for patients.

A student working in this environment will become familiar with medical imaging analysis, safety engineering, treatment planning systems. Example projects include:

• AI translation and applications in radiotherapy: Trainees would participate in developing and translating AI or other computation models to assist radiotherapy, such as segmentation and automatic treatment planning.
• Independent AI quality assurance in prospective workflow: Trainees would participate in building independent safety layers or QA tools that screens AI results, such as segmentations, before they influence patients.
• Online adaptive radiotherapy: Trainees would participate in designing decision-support systems that determine who/when/how to adapt; and quantify throughput and dosimetric trade-offs across adaptation strategies using clinical information, imaging features and dose-based evidence.

PubMed

Google Scholar

Education & Training

PhD, University of Kentucky, 2023