Segmentation of brain lesions in diffusion-weighted MRI using thresholding technique

This paper presents brain lesion segmentation of diffusion-weighted magnetic resonance images (DW-MRI or DWI) based on thresholding technique. The lesions are solid tumour, acute infarction, haemorrhage, and abscess. Preprocessing is applied to the DWI for normalization, background removal and enhancement. Two different techniques which are Gamma-law transformation and contrast stretching are applied for the enhancement. For the image segmentation process, the DWI is divided by 8×8 regions. Then image histogram is calculated at each region to find the maximum number of pixels for each intensity level. The optimal threshold is determined by comparing normal and lesion regions. By using Gamma-law transformation, 0.48 is found as the optimal thresholding value whereas 0.28 for the contrast stretching. The proposed technique has been validated by using area overlap (AO), false positive rate (FPR), and false negative rate (FNR). Thresholding with gamma-law transformation algorithm provides better segmentation results with AO, FPR, FNR (0.68, 0.14, 0.18) compared to contrast stretching (0.62, 0.15, 0.23).