Auto Multi-ROI Detection on Medical Images for Data Hiding using Improved ORB Features and Optimised Clustering Algorithms

Abstract

Telemedicine plays a vivacious role in collaborative clinical decision making, however secure and scalable data sharing mechanism is a challenge. The Office of the National Coordinator for Health Information Technology’s (ONC) has given a roadmap for data protection and regulation that telemedicine architecture should abide. To address the confidentiality problem, various encryption schemes are used along with several steganography methods to hide the clinical data in respective medical images. By doing so, Integrity and confidentiality of these data can be maintained concerning telemedicine applications. Several pieces of research were done to embedded Electronic Patient Record(EPR) into the Non-Region-of-Interest (NROI) segment of that medical image, of which ROI is selected by a practitioner.  The Paper proposes two distinct mechanisms for auto multi Region-of-Interest detection which can be used for any EPR embedding purposes. Auto ROI detection is done by extracting feature points from the medical image using improved Oriented FAST and Rotated BRIEF (ORB) and ROI framing using controlled K-Means and HDBSCAN clustering algorithms. The experimental result shows that the proposed methods are efficient to the all type of medical image for Auto RONI detection.