An Efficient and Robust Feature Extraction Based Biomedical Image Retrieval and Classification Model
Recently, there is an exponential growth in the number of medical images being generated from hospitals and medical organizations. Effective utilization of the images saved in the database for image retrieval and classification process becomes highly essential to accurately diagnose the patient and for further examination. In this view, this study presents new medical image retrieval and classification (MIRC) model, which retrieves the medical images related to the query image (QI) from the database and classifies the images. The proposed MIRC model involves three main processes namely feature extraction, image retrieval and image classification. At the initial stage, a set of two features namely texture and shape features get extracted from the input image. Then, the retrieval of the images takes place using a Euclidean distance based similarity measure. In line with, multilayer perceptron (MLP) based classification model is applied to classify the retrieved images. To improve the training process of MLP, hybridization of particle swarm optimization (PSO) and genetic algorithm (GA) is derived. During the testing phase, the proposed model determines the feature vectors of the test image and computes the similarity measure for retrieving the related images. Finally, the retrieved images undergo classification and allocate class labels to every test image. For validation, a set of benchmark NEMA CT images are employed. The experimental outcome clearly stated that the presented model has offered effective retrieval and classification performance by attaining maximum precision of 80.84%, recall of 86.49% and accuracy of 84.13%.