A framework for performing textural analysis and classification of prostate ultrasound images.

摘要

This thesis presents an integrated framework for performing textural analysis and classification of transrectal ultrasound images of the prostate into clusters potentially representing different tissue areas. The approach is based on the textural feature analysis proposed by Haralick [1] and the Minimum Squared Error classification algorithm [2]. The Java Textural Analysis/Classification (JTAC) application developed as part of this thesis offers significant reduction in run time, potentially allowing more accurate, objective diagnoses to be performed within clinical settings, and allows the investigation of parameters associated with textural and classification processes. The textural analysis algorithms focuses on five of the fourteen features proposed by Haralick including Angular Second Moment, Contrast, Inverse Difference Moment, Entropy, and Sum Entropy. Using this integrated approach, specific results for several cases are presented and general conclusions are developed. The approaches implemented in this framework are outlined in this thesis as well as improvements and areas of future investigation.