Using block-based multiparameter representation to detect tumor features on T2-weighted brain MRI images

The objective of this paper is to present an analytical method for digitized images to detect tumors or lesions in a medical decision support system.

The authors have developed a simple method of tumor detection using three parameter values: edge (E), gray (G), and contrast (H). The method proposed here first studied the VHD (Visible Human Dataset) input brain feature using EGH parameters that divided the input image into fixed-size blocks (templates). The EGH parameters for the feature blocks were calculated and parameterized to detect the occurrences of abnormalities. These abnormal blocks were then marked for interpretation.

Measurements of the following medical dataset were performed: 1) different time-interval images from the same dataset, 2) different brain disease images from multiple datasets, and 3) multiple slice images from multiple datasets. Our experimental results illustrate the ability of our proposed technique to detect tumor blocks with conceptual simplicity and computational efficiency.

In this paper, we present output examples from our prototype system, comparing detection accuracy and system performance.