A Novel Optimal Deep Learning Based Image Retrieval and Classification Model for Biomedical Images
In recent times, there is an exponential growth in the generation and utilization of medical images, which offers extensive details about the anatomical structure of a patient. Therefore, medicinal images have been utilized for diagnostics and research purposes to understand the deep insight into the reason and treatment of diverse diseases. To retrieve and classify the medicinal images from huge databases, it is needed to develop an effective medical image retrieval and classification model. In this paper, a new medical image retrieval and classification model has been developed incorporating three different processes namely feature extraction, similarity measurement based image retrieval and image classification. At the earlier level, texture and shape features are extracted from the original image. On the application of new query image as input, the image retrieval process is executed utilizing a Euclidean distance based similarity measure to retrieve the relevant images. Then, grey wolf optimization (GWO) tuned deep neural network (DNN) called GWO-DNN model is applied for classification task. The hyperparameter tuning of DNN model takes place by the use of GWO algorithm. Finally, the classification process gets executed and assign class label to the applied test image. To validate the results, a benchmark NEMA CT images is utilized. The experimental results clearly portrayed the significance of the proposed model by attaining a maximum precision of 89.39%, recall of 94.18% and accuracy of 93.73%.