Shannon’s entropy and watershed algorithm based technique to inspect ischemic stroke wound

Abstract

Ischemic Stroke (IS) is usually initiated due to the neurological shortfall in human brain and which can be recognized by inspecting the periphery of the brain sections. In this paper, a two-step procedure is proposed to extract and evaluate the IS injury from brain Magnetic Resonance Image (MRI). In the initial step, Social Group Optimization and Shannon’s entropy based tri-level thresholding is executed to enhance the IS section of the test image. During the second step, enhanced IS section is then mined using the marker-controlled Watershed (WS) algorithm. The proposed practice is tested on benchmark ISLES 2015 dataset. Performance of the WS segmentation is also verified with the segmentation approaches, like seed-based region growing (SRG) and the Markov Random Field (MRF). The outcome of this study authenticates that, WS provides enhanced picture likeness indices, like Jaccard (90.34%), Dice (94.92%), FPR (7.77%) and FNR (2.65%) compared with SRG and MRF.