Indian Journal of Science and Technology
Year: 2016, Volume: 9, Issue: 48, Pages: 1-9
K. Sudharani1*, T. C. Sarma2 and K. Satya Prasad3
1 VNR Vignana Jyothi IET, Hyderabad - 500090, Telangana, India; [email protected]
2 NRSA, Hyderabad - 500037, Telangana, India; [email protected]
3 JNTU, Kakinada - 533003, Andhra Pradesh, India
*Author for correspondence
VNR Vignana Jyothi IET
Background: A tumor is a gathering of tissues that grows in a disordered manner that normalizes growth. Brain tumor detection in MRI, CT, PET scan is most interesting area in the medical image field. The main objective is developing a novel technique i.e., histogram based region related detection of brain tumor. Method: An automatic algorithm for detection of brain tumor and its tumor segmentation using MRI T1 weighted, MRI flair images is presented. The proposed algorithm divides the brain into four regions Top half, bottom half, right-side half and left side half. It utilizes pixel intensity levels obtained from the each region histogram of an image for the segmentation as the result is more useful to analyze the raw image. The mathematical descriptions like statistical parameters of the proposed approach are presented in detail. The proposed algorithm reduces misclassification errors where the minimal dissimilarity within each object by its own cannot guarantee the desirable result and a comparison is made with the existing techniques like entropy and moments thresholding. Findings: Brain tumor is effectively detected and located in the brain by dividing the whole brain image into four regions. After dividing the brain into four halves histogram is applied individual part of the divided region. The histogram is the no of the amount of the pixel intensity. A performance evaluation is also done by checked the results by reference/ground truth MRI images through which sensitivity and accuracy of the proposed algorithm can be determined. The performance measures Sensitivity, Specificity, Accuracy and Similarity index obtained from the proposed method are 91.429%, 86.667%, 90%, 92.754%, respectively. The statistical parameters reveal the algorithm stability and reliability. The results obtained by this algorithm are included in the study and found comparatively better than the results obtained with Entropy and Moments Thresholding techniques. Applications: The proposed histogram based brain tumor detection and analysis efficiently dealt with detection of brain tumor and image classification procedure. Doctors practice the information obtained from the algorithm results to verify the most suitable course of treatment.
Keywords: Entropy, Histogram, Morphology, Moments, Segmentation
Subscribe now for latest articles and news.