Optimal Feature Subset Selection and MSVM Classification Based CBIR for Medical Images
International Journal of Advanced Trends in Computer Science and Engineering
The main objective of this study is to design and implement a novel CBIR system for medical image retrieval system by comparing the features of the query image with the DB image. Here it is discussed about the feature selection for image comparison and retrieval. At present the entire process uses a wrapper approach using Association rules-based on feature selection and PS-classifier. As a result of this experiment, the proposed model is comparable to the other models on diabetic retinal
... s. Here it is introduced three different classifiers such as PS-Classifier, Artificial Neural Network and Association rules for selecting an optimal feature subset on diabetic retinal dataset as DRIVE to evaluate the effectiveness of proposed feature selection method. From the optimal feature subset, the Multi Class Support Vector Machine algorithm is used for classifying the abnormal condition of the medical images. The entire system is experimented in MATLAB software and the results are verified. From the results, it is observed that the feature selection based CBIR outperforms than the other existing one discussed in the literature. This paper obtained an improved accuracy in all classifiers in feature selection for any DR Dataset. Also, the result shows that the feature selection based CBIR has more efficiency in terms of relevant image retrieval. In addition, the specificity and sensitivity also improved after feature selection.