Abstract: Artificial intelligence (AI) has demonstrated its tremendous power in recent years, ranging from face recognition to autopiloting cars. Its applications in medicine have also achieved remarkable successes in a variety of problems. At the University of Texas Southwestern Medical Center, a group of medical physicists are dedicated to the development and deployment of AI technologies to solve problems in radiotherapy. This presentation will first give an overview of our AI-related researches in the Medical Artificial Intelligence and Automation lab. It will then focus on studies in Dr. Xun Jia’s lab using AI techniques to mimic human behaviors for problem solving. While the mainstream of current AI researches is to use deep neural networks to establish correlations between input and output data, we focus our studies on  developing novel techniques to generate human-like behaviors that can intelligently respond to environments in the problem-solving processes. Specifically, solutions to two challenging but representative problems in radiotherapy will be presented. In the problem of sigmoid colon segmentation, we developed an iterative deep-learning segmentation approach that imitated a human to segment the sigmoid colon slice by slice, while considering the connectivity of the organ between neighboring slices. This approach achieved the state-of-the-art result with a Dice similar coefficient of 0.84. For the automatic treatment planning problem, we developed a virtual treatment planner trained via end-to-end deep reinforcement learning. The virtual planner spontaneously learnt to generate intelligent treatment planning behaviors and can operate an optimization engine to develop high-quality plans. We demonstrated the feasibility and potential of this approach in both external beam radiotherapy for prostate cancer and brachytherapy for cervical cancer. The second half of this talk will discuss challenges when using AI to solve problems in radiotherapy. We will present the requirements to train the underlying deep neural networks, challenges to meet these requirements, and other mathematical and practical issues to consider. We hope the presentation will stimulate discussions and facilitate the effective use of AI techniques to advance radiotherapy.