A database with physician input & novel eye tracking techniques designed by an RIT ProfessorSomeday, digital archives of stored biomedical images might make it possible for the doctors to easily access critical information within seconds. However, the current reality isn’t really up to speed as such and databases have been overwhelmed virtually by medical imaging explosion. Professor Anne Haake of Rochester Institute of Technology won grants recently from the National Institutes of Health and National Science Foundation to address the problem. She envisions creation of an image database with the intended end uses providing the input and designed with very flexible user interfaces from the beginning. Together with her interdisciplinary team, Haake will be developing a prototype with inputs sourced from dermatologists for refining the mechanism used in searching for images of different skin conditions.

Haake says that it is important that users are involved right from the beginning. This is true especially in the field of biomedical where a lot of domain knowledge exists and will be used specifically for each particular specialty. Haake understands too well why it is necessary to make the biomedical images useful. Before pursuing computing & biomedical informatics, she launched her career as a professional developmental biologist. As such, this project is a combination of both her strengths and the research she conducted at NIH National Library of Medicine while on sabbatical was a great inspiration for her.

Dr. Cara is assisting with image descriptions

For this project, Dr. Cara Calvelli, a professor and a dermatologist at RIT’s College of Science in Physician Assistant program has recruited residents, PA students and dermatologists. In addition, she is also assisting in describing the sample images properly and some of them are actually sourced from her large collection. Calvelli says that the best way of learning is seeing patients with various disorders again and again. If you are not able to get patients themselves, the second best option is to get good images and learn how to give proper descriptions for them.

How the funding Haake won will be used

The funding won by Haake is meant to support her visual perception research by use of eye tracking and design of content based system for image retrieval accessible via gaze, touch, gesture and voice. The portion of NIH in the project is going to be used for fusing medical knowledge and image understanding. The researchers who are working on the content based image retrieval are based with the challenge of bridging the ‘semantic gap’. There is a possibility of the search functions going awry when the computer engineered algorithms has tripped on the nuances and fails to easily distinguish between the disparate objects like a ship and a whale. However, semantic hiccups can be prevented from occurring by designing a system based largely on the knowledge of end users.