Brain tumor detection using self-adaptive K-means clustering

Brain tumor detection is an important diagnostic process in medical field. Magnetic resonance imaging (MRI) is the prime imaging technique while analysing the brain/skull with respect to brain tumor localization and detection. The brain MRI images show a complex network of brain cells along with bony structures and suspected solid growth if present. Therefore, in order to extract the growth, a segmentation process is required. In original K-means algorithm, the no. of clusters are define by the user i.e. user input is required. However, this limitation is overcome by using the self-adaptive K-means clustering algorithm to detect brain tumor accurately and in minimal execution time. A sobel edge detection technique is followed to extract the edges of the segmented brain tumor from its surroundings. In self-adaptive k-means clustering, the number of clusters are computed by computing the peaks in histogram. The segmented part is then processed to binary image format for its size and location estimation. The gray version is used to extract textural and color based features for nature of growth analysis. The final segmented part is applied the size estimation algorithm for tumour area and perimeter estimation.