Image Search Engine

Unmet Need
When treating patients, doctors primarily perform a visual examination aided by reference materials and photo libraries. This makes the process of identifying unknown conditions slow, painful, and subject to trial and error, often resulting in unnecessary pain and cost to the patient. While massive online medical photo libraries exist, they are functionally limited by organizational structures, reliance on Boolean search, and other archaic methodologies. These systems are anachronistic and force an overreliance on doctors’ primary knowledge, limiting patient access to the best possible care.
 
Technology Overview
Through the use of an image based search engine, unknown lesions, sores, physical symptoms, and MRI/CT imagery can be photographed and almost instantly compared to millions of similar images to find like cases. The system first compares an input image to other images in a database, determining the extent of linear transformation required to match the input image with the database images, and then pulls the database images representing the closest match. This system is highly effective for comparing aspects of physics sores such as length, width, size, and general shape. For larger, more complex images displaying a complicated sore or lesion, the system divides the image into a matrix of hundreds of sub-images, performing the first process on each shard. Not only does this allow the engine to account for substantially more complex variables (e.g. curvature, dynamics, depth…), it allows the system to run hundreds of comparisons simultaneously, substantially reducing the time necessary to search for large, complex, multivariable images. This technology has both clinical and consumer applications.
 
Stage of Development
Proof of concept, early stage prototype.

Category(s):