Embracing AI-Assisted Programming for Problem Solving and Development in Medical Imaging Physics
Description
As artificial intelligence rapidly transforms how software and tools are developed, medical physicists are increasingly exploring AI-assisted programming -- using models such as ChatGPT (and Codex) and Claude -- to accelerate problem solving and prototyping in imaging physics. This educational symposium will demonstrate how large language models (LLMs) can serve as intelligent collaborators in developing solutions for practical challenges in medical imaging physics, such as quality control automation, data analysis pipelines, MR safety systems, and protocol management. Speakers will share real-world experiences integrating AI-assisted coding environments (e.g., ChatGPT or Claude in VS Code) into their workflows, highlighting both the productivity gains and common pitfalls. For example, one project leveraged LLMs to develop a MATLAB-based MRI ACR geometric accuracy test program within days, later translated into Python -- illustrating both the power and limitations of AI-assisted cross-language translation. Another project used Claude to develop a comprehensive imaging QC tracking system that includes ACR accreditation tracking, personnel credential management, and automated DICOM analysis, as well as an MR safety training dashboard for staff and students. By combining diverse case studies and practical demonstrations, this symposium will help participants move beyond curiosity toward competence in AI-augmented development and foster a sustainable approach to innovation and informatics literacy in the evolving landscape of medical imaging physics.