Abstract:

BACKGROUND: Due to the low resolution of naked eyes, many small details and texture changes in morphology are not easy to be found, it will affect the early judgement of diseases.
OBJECTIVE: To provide a new way of thinking and methods for the computer diagnosis of diseases in digital medical images, which helps doctors to detect and diagnose the early malignant lesions and improve diagnostic efficiency and accuracy.
METHODS: Outlier data mining technique was used to analyze large data sets, extract the hidden, unnoticed, and easily discarded, but very useful information, and find out the rules and patterns of medical diagnosis to assist doctors to diagnose disease.
RESULTS AND CONCLUSION: Experiments show that outlier analysis algorithm based on clustering of the medical image pixels is feasible for the brain lesions.