ExPose: A Web-based Platform for AI-assisted Generation of Exercise Illustrations for Rehabilitation Prescriptions

The exercise illustration of rehabilitation prescriptions is to provide a more intuitive and understandable way to convey medical guidance through visual means. Currently, the generation of exercise illustrations for rehabilitation prescriptions relies mainly on manual drawing, either using software or by hand. However, this method suffers from not only complexity and high labor costs, but also low production efficiency. While today's Generative AI (GenAI) has achieved great success in text and image generation, images that meet descriptions can be generated quickly from prompts or images using diffusion models and the ControlNet model. However, WebUI and ComfyUI, the most popular user interfaces for diffusion models at present, are too general for the specific task of generating exercise illustrations from rehabilitation prescriptions, partly due to the application and domain-specific complexity. This research proposes ExPose, a dedicated web-based platform, for the generation of exercise illustrations from rehabilitation prescriptions. With embedded image editing functions, ExPose enables users to crop, modify, and annotate the generated intermediate images, thus providing streamlined image processing and facilitating iterative image retouching and enhancement. The use cases demonstrate that ExPose generates high-quality exercise illustrations that better meet expectations from rehabilitation prescriptions. Evaluation results show that ExPose achieves a high rate of user satisfaction.