CANCER PREVENTION INITIATIVE: AN INTELLIGENT APPROACH FOR THYROID CANCER TYPE DIAGNOSTICS

Abstract

In recent, medical Image mining has witnessed to be one of the emerging fields of machine learning.  Particularly; the classification problem of DICOM (Digital Imaging and Communications in Medicine) images has become a prominent challenge. Thyroid cancer must be detected as earlier as possible; a little delay would extremely be proved hazards for human health and may be resulted into the most fatal threat to human life. Infect in-depth study of physical components of cells of FNAB (Fine needle aspiration Biopsy) would help to refine the results and provide more precise decisions about the potential occurrences of cancer, this paper proposes a system, so called ‘TCTD’ (Thyroid Cancer Type Diagnostics), which aims to assist the doctors during their diagnostic process conducted for thyroid follicular carcinoma and its sub-types. There are five main steps of our methodology. In first step image pre-processing techniques are used. In the second step; we use ensemble methods, such as Multi-SVM to build decision model and to analyze the frequencies between the variables. In third, fourth and fifth steps, model testing, result visualization and performance evaluation are performed by using precision, recall and f-measure estimations. The measured classification accuracy of proposed system is about 98.50% using 10 K-fold cross validation.