VISUAL NATURAL LANGUAGE PROCESSING OF MEDICAL IMAGES FOR ENAHANCED VALUE

Abstract

Proper analysis of histological images is critical for medical professionals because they can be used to determine the presence of diseased tissues. However, the process can be time-consuming and subject to human error. The Arizona Center for Accelerated Biomedical Innovation desires a low cost, unique solution that is capable of deconstructing a medical histological image to extract relevant information for classification.The design team’s solution is a computer application, FractalEyes, that allows medical professionals to select a histological image and output an image separated by a grid that is characterized by color frequency. The concept of this application is derived from Natural Language Processing, where sentences are deconstructed to enable computers to contextualize and understand language. FractalEyes segments an image into uniform sections termed “image voxels” -similar to how a sentence is divided into words. Color information from each division is analyzed to generate a color frequency. The relation of the color frequency to each grid section provides context to the image, thereby allowing a program to classify types of images. The FractalEyes software is comprised of five subsystems that interact in sequential order. The subsystems are 1. GUI 2. Preprocessing 3. Feature Extraction and Comparison 4. Patient Data Processing 5. Storage.