MEDICAL IMAGE FEATURES EXTRACTION USING THE WAVELET TRANSFORM

Abstract

Technique of dividing voxels into 3D sections (sub volumes) that correspond to significant physical things is known as 3D volume segmentation. This makes the data more understandable, easier to analyze, and useful for future applications. Multiresolution analysis, or MRA, makes it possible to preserve a picture based on specific blurring or resolution levels. Wavelets have been used for image compression, demising, and classification due to their multiresolution quality. The application of effective medical volume segmentation strategies is the main topic of this research. Feature extraction has been done using multiresolution analysis, which includes 3D wavelet and ridge let. The volume slices can be segmented using Hidden Markov Models (HMMs). The Region of Interest (ROI) can be accurately detected using 3D procedures, according to a comparison study that was conducted to assess 2D and 3D approaches.
The goal of the experimental investigation described in this work is to create an automatic image segmentation system that can classify ROI in medical pictures that are taken by several medical scanners, including PET, CT, and MRI.The suggested segmentation system makes use of multiresolution analysis (MRA) with wavelet, ridge let, and curve let transforms. Classifying malignancies in human organs using shape or gray-level information from scanner output is especially difficult since soft tissues' gray-level intensity overlaps and organ shapes vary across different slices in the medical stack.
A novel expansion of wavelet and ridge let transforms, the curve let transform seeks to address intriguing phenomena that arise along curves.