Application Of Supervised Machine Learning To Characterize Brain Tissue And To Discriminate Benign Lesions, Various Grades Of Glioma And Metastasis

Abstract

Supervised Machine Learning (SML) an extremely powerful classifier was
applied for diagnosing the various pathological lesions in the brain, like edema, multiple
sclerosis MS), glioma of different grades and metastasis. MR Images may show structural
changes in the brain lesions (Figure 1). MR Spectroscopy can also show change in the
metabolite peaks and quantities in different disease state
(Figure 2). But it is frequently difficult to diagnose the exact disease. Use of SML by two
strategies like Artificial Neural Network (ANN) and Support Vector Machine (SVM) helps
identifying the condition in doubtful cases. The SVM and ANN train on data sets gathered from
different patients based on input variables – Refractive Index,T2 relaxation values, Choline
(CHO), Apparent Diffusion Coefficient (ADC), Creatine (CR), CHO/NAA (N-acetyl aspartate),
CR/NAA, LIP/LAC (Lipid/lactate), MI ( Myoinositol), CHO/CR and T2 value in the periphery
of lesion. Refractive index is a vital physical parameter. After training the data, prediction by
ANN and SVM show high accuracy in diagnosis. The training and testing have been carried
out by Neural Tool in ANN and SVM classifier tool in MATLAB respectively.