Cross Entropy Based Thresholding Segmentation of Magnetic Resonance Prostatic Images Using Metaheuristic Algorithms

Abstract

Magnetic resonance images are relevant sources of information for detecting and diagnosing a large number of illnesses and abnormalities. Segmentation of digital images helps to classify the pixels in the different regions according to their intensity level. Segmentation of digital images implemented on magnetic resonance images can help experts to improve the performance of evaluations and make a correct differential diagnosis to indicate the appropriate treatment. This chapter proposes the use of MFO, SCA and SFO algorithms to search for 2, 3, 4, 5, 8, 16 and 32 threshold values using minimum cross entropy function to segment prostatic magnetic resonance images, and implementing statistical metrics like PSNR, SSIM, and FSIM to measure the quality of a segmented image, quantified and establish the proper comparison frame, so visual comparison is not enough. The approach is tested on a set of benchmark images to demonstrate that the segmentation of digital images can improve the detection of prostatic abnormalities or illnesses.

Appendix: MRIs Detailed Series Header Information

• sh.ExamDescription = ‘PROSTATE STAGING’;

• sh.SeriesDescription = ‘T1 AXIAL (COIL OUT) 2 STACKS’;

• sh.SequenceName = ‘MEMP’;

• sh.PatientAge = 60;

• sh.PatientWeight = 80,000;

• sh.FieldStrength = 15,000;

• sh.SeriesDate = ‘Day 1′;

• sh.RepetitionTime = 700,000;

• sh.EchoTime = 8000;

• sh.Excitations = 2;

• sh.FlipAngle = 90;

• sh.FrequencyDirection = ‘Row’;

• sh.SliceThickness = 5;

• sh.MatrixSize = [256 256];

• sh.PixelSize = [1.093750 1.093750];

• sh.Fov = [280 280];

• sh.PixelValueOffset = 0;

• sh.NumberOfImages = 46;

• sh.Gamma = 1;

• sh.ImagePositionPatient1 = [−139.453125 − 156.053131 211.800003];

• sh.ImageOrientationPatient1 = [1 0 0 0 1 0];

• sh.ImagePositionPatientN = [−139.453125 − 156.053131 − 58.200001];

• sh.ImageOrientationPatientN = [1 0 0 0 1 0];