A Multiparameter Hierarchical Representation Using Region-Based Estimation Model for Detecting Tumor in T2-Weighted MRI Brain Images

Abstract

The objective of this paper is to present an analysis method for digitised medical images to allow diagnosis and interpretation for medical practitioners or medical students. A simple diagnosis method was developed using multiparameter value; edge (E), gray (G), and contrast (H), to develop a decision support system for interpreting and analysing T2-weighted MRI brain images. The hierarchical representation proposed here is able to study the sub-regions EGH parameter in comparison with the estimation model for abnormal occurrences, and depending on the regions of abnormality, diagnosis and prescription is provided. In this paper, three different experiment sets are introduced to study the proposed method: 1) Set I: different time interval images, 2) Set II: different brain disease images, and 3) Set III: multiple slice images from different age and gender. Experimental results illustrates that our proposed technique here incurred an overall error smaller in comparison with our previous proposed method. In particular, the proposed method allowed decrements of false alarm and missed alarm, which demonstrates the effectiveness of our proposed technique. In this paper, we also present a prototype system, known as PCB, to evaluate the performance, accuracy and robustness of the algorithms and reveal the proposed approach capabilities in interpretation and diagnosis.