Vision-based AI is transforming digital health by integrating cutting-edge imaging technologies with advanced analytical methods, resulting in significant improvements in diagnostic precision and patient care. These technologies enable early disease detection by identifying subtle anomalies in medical images, streamline clinical workflows, and accelerate research by processing large datasets efficiently.
In addition, the integration of large language models (LLMs) with vision-based AI—forming Vision LLM systems—further enhances healthcare applications by combining visual data interpretation with natural language processing. This multimodal approach not only improves the generation of clinical reports and supports decision-making but also opens new avenues for comprehensive patient care.
The workshop will cover a broad spectrum of topics that demonstrate the innovative applications of vision-based AI in digital health.
Accepted workshop papers will be published in the official ICCV workshop proceedings.
We invite paper submissions with topics include, but not limited to:
Vision LLM for Healthcare: This topic explores the integration of large language models (LLMs) with vision-based AI to enhance healthcare applications. It will address how vision LLM systems can interpret medical images, generate detailed clinical reports, and support decision-making by combining visual and textual data. Participants will learn about innovative approaches that leverage multimodal data, discuss the challenges of aligning language understanding with visual context, and examine case studies where vision LLMs improve diagnostic accuracy and treatment planning in clinical settings.
Medical Image Analysis and Diagnostics: Examining how AI-driven techniques analyze various imaging modalities to detect abnormalities, improve diagnostic precision, and support treatment planning.
Real-Time 3D Reconstruction for Medical Endoscopy: This topic offers a concise overview of current 3D reconstruction methods for medical endoscopy, addressing challenges like featureless surfaces, variable lighting, and deformable structures. It examines techniques from monocular to multiocular approaches, compares active and passive methods, and discusses adaptations for both flexible and non-flexible endoscopes. The session will also review key error metrics, hardware considerations (GPU vs. FPGA), and benchmarking using datasets such as KITTI and Middlebury.
Applications in Digital Health: This topic explores the diverse ways vision-based AI is impacting digital health beyond traditional diagnostics. It covers innovative applications such as skin cancer detection, where imaging is used to identify cancerous lesions at early stages; wound monitoring through sequential image analysis to track healing progress; gait analysis via video-based methods to assess patient mobility; and remote patient monitoring using live video streams to continuously track vital signs and patient behaviors.
Paper formatting: Papers must be a minimum of 4 pages and may not exceed 8 pages, including all figures and tables, formatted in the official ICCV style. Additional pages are permitted only for references. Please download the ICCV 2025 Author Kit for detailed formatting instructions.
Please follow the ICCV 2025 Author Guidelines and submit your paper through the VADH 2025 submission portal on Openreview.
Accepted papers will be published in the ICCV 2025 Workshop Proceedings following the ICCV 2025 publication guidelines.
Important Dates
Paper Submission: June 27
Notification to authors: July 9
Camera-ready submission: August 18, 11:59 PM, Pacific Daylight Time
Workshop: October 19