Content Based Medical Image Retrieval Using Clustering Technique

Abstract

Tumors of the brain are life-threatening and life-defying diseases for humans. For the radiologist, finding a suitable MRI brain tumour picture might be a difficult challenge. Text-based techniques are still the most common method used by most search engines to retrieve pictures. The major issue in MRI image analysis is that the MRI machine captures low-level visual information and the assessor identifies high-level data. To bridge this conceptual chasm, researchers in this study devised a brand-new feature extraction method. CBMIR (Content-Based Medical Image Retrieval) is a new approach to finding pictures of brain tumours in vast databases. MRI pictures are first cleaned up by using a variety of filtering procedures. This is followed by the construction of an algorithm for extracting representative features from MRI images using the Gabor filtering approach and the Walsh-Hadamard transform (WHT) technique, both of which concentrate on a certain frequency content at a specific area of the picture. Using Fuzzy C-Means clustering Minkowski distance metrics, we can then get the most accurate and dependable picture by comparing our query image to a database of photos. Brain tumour MRI images were used to evaluate the suggested technique design. Experiments on brain tumour detection using our suggested method show that it outperforms most of the current methods including Gabor, wavelet, and Hough transform. For radiologists and technicians, the suggested technique will help to construct an automated decision support system that will deliver repeatable and objective outcomes with high accuracy.