Visual Natural Language Processing of Medical Images for Enhanced Value

Medical imaging is a critical component of disease diagnosis in medical practice. Although health care providers have used advances in medical imaging technologies to implement more effective point-of-care strategies, the analysis of medical images remains inefficient and highly subjective. The solution developed is a full image classification, feature extraction and feature analysis tool called FractalEyes. The system can classify an input image, perform application-specific feature extraction and analyze the features to provide health care providers with a more quantitative measure of diagnosis. The software uses a mutual information classification algorithm from the scikit-learn Python library. The program communicates both graphically and quantitatively which features of the image are useful in the image’s classification. The team used this proof-of-concept system to classify histological images of human cells. The idea can be extended to applications ranging from disease diagnosis to counterfeit item detection.