SEGMENTATION OF STROKE REGIONS FROM DWI AND ADC SEQUENCES USING A MODIFIED WATERSHED METHOD

Ravi S, A.M. Khan

Abstract: Stroke is a cerebro-vascular accident that alters the supply of blood to the brain cells. Stroke is second highest cause for morbidity rate among humans after cardiovascular arrest. Stroke can hardly be observed in T1 and T2 sequences of MRI on the onset. The doctors usually prefer diffusion weighted imaging sequence and apparent diffusion coefficient mapping to diagnose strokes. Segmentation is the necessary part of medical image processing to get more information on the abnormal regions. Watershed method to segment these regions is one of the widely used techniques. The watershed algorithm is sensitive to minor changes in the regions. This sensitivity and the presence of noisy pixels make the algorithm to suffer from over-segmentation. This paper presents an effective pre-segmentation approach to minimise over-segmentation by segmenting the affected region from the normal tissues.

Keywords: Stroke, MRI, Segmentation, Watershed Algorithm, DWI, ADC

DOI: https://doi.org/10.15623/ijret.2015.0424001