Detecting tumours by segmenting MRI images using transformed differential evolution algorithm with Kapur’s thresholding

Abstract

The speed and accuracy with which the patient affected with brain tumour is diagnosed and monitored, plays a very crucial role in providing treatment to the patient. During the diagnosis of the diseased part, a constant demand is anticipated to easily extract the specific region of interest within the complex medical image. This task of extracting only the diseased portion amid the complex body parts in the complex medical image can be achieved by image segmentation. Accuracy and speed of extracting the points or area of interest within the multipart medical image can be improved by using various evolutionary techniques. Differential evolution (DE) is an efficient evolutionary technique that can be used for solving optimisation problem like image segmentation. The main disadvantage of classical evolutionary technique is its inability to adapt its solution algorithm to a given problem. Owing to this need, more adaptable and flexible algorithms are in demand. Numerous variants of DE exist which differ in their solutions. Here, a variant of differential evolution named as transformed differential evolution (TDE) is presented which has an improved mutation strategy that is optimised to fewer function evaluations. This variant is combined with the Kapur’s multi-level thresholding for segmenting magnetic resonance imaging (MRI) images and to extract only the regions of tumour. The results obtained using TDE with Kapur’s multi-level thresholding were compared with the results using traditional Kapur’s technique and the new results improved profoundly. By introducing TDE in multilevel thresholding, the computational time significantly reduced and the resultant image quality improved greatly.