Pseudo-colour with K-means clustering algorithm for acute ischemic stroke lesion segmentation in brain MRI
Document Type
Article
Publication Date
4-1-2021
Abstract
Segmentation of an acute ischemic stroke from a single modality of a greyscale magnetic resonance imaging (MRI) is an essential and challenging task. Recently, there are several numbers of related works on the automatic segmentation of infarct lesion from the input image and give a high accuracy in extraction of infarct lesion. Still, limited works have been reported in isolating the penumbra tissues and infarct core separately. The segmentation of the penumbra tissues is necessary because that region has the potential to recover. This paper presented an automated segmentation algorithm on diffusion-weighted magnetic resonance imaging (DW-MRI) image utilizing pseudo-colour conversion and K-means clustering techniques. A greyscale image contains only intensity information and often misdiagnosed due to overlap intensity of an image. Colourization is the method of adding colours to greyscale images which allocate luminance or intensity for red, green, and blue channels. The greyscale image is converted to pseudo-colour is to intensify the visual perception and deliver more information. Then, the algorithm segments the region of interest (ROI) using K-means clustering. The result shows the potential of automated segmentation to differentiate between the healthy and lesion tissues with 90.08% in accuracy and 0.89 in dice coefficient. The development of an automated segmentation algorithm was successfully achieved by entirely depending on the computer with minimal interaction.
Keywords
Acute ischemic stroke, Clustering, MRI, Pseudo-colour, Segmentation
Divisions
fac_med
Funders
Yayasan Sarawak through the Yayasan Tun Taib scholarship,Universiti Malaysia Sarawak (UNIMAS) through the Yayasan Tun Taib scholarship
Publication Title
Pertanika Journal of Science and Technology
Volume
29
Issue
2
Publisher
Universiti Putra Malaysia