Performance Analysis of Ischemic Stroke Lesion Segmentation in Brain MR Images using Histogram based Filter Enhanced FCM

A medical condition which threatens life by blocking brain’s blood supplying arteries is called ischemic stroke. An early and accurate diagnosis of brain stroke is possible through evaluation of different modalities (DWI, FLAIR Tl, T2) of medical image format available in Magnetic Resonance imaging (MRD. Delineation of Stroke lesion from whole brain MRI, manually by Expert radiologists and neurologist consumes larger proportion of time and subject to intra- and inter-observer variation. Hence there is a demanding need of computer based automated algorithm which extracts out the stroke lesions from the whole brain MRI and accelerate the process of early stroke diagnosis. With the above objective this work proposes Histogram Based Filter Enhanced Fuzzy C Means that brings in the pro role played by Histogram (unique Grey level) wise clustering rather than pixel wise clustering as followed in classical FCM. The objective function of Histogram based FCM is smoothened further by filtering through the median filter which incorporates spatial information into the FCM clustering technique where local information alone is considered. Above Changes in classical Fuzzy C Means Clustering technique through Histogram and median filters provides a greater improvement with maximum value of 99% accuracy in DWI and FLAIR 79% Dice in DWI, 9S% precision in DWI, SS % sensitivity in DWI when compared with available state of art methods standard FCM, FCM S1, FCM S2 and EnFCM.