Enhancing Liver MRI Image Classification through an Innovative Hybrid Approach of Genetic and Social Spider Techniques for Effective Feature Selection

Abstract

One of the vital organs in the human body is the liver. A variety of diseases brought on by the organs' poor functioning every day include cirrhosis, hepatitis, and fatty liver. Many people's irregular eating habits, intake of alcohol, etc., are contributing to their development of liver illnesses. When it comes to the early identification and detection of liver problems, medical imaging is essential. Nonetheless, a major issue is accurately identifying essential features from medical imaging. The effectiveness of optimization methods inspired by nature is investigated in this study regarding feature selection for hepatic medical imaging. In order to determine these strategies' applicability and efficacy in improving diagnostic accuracy, the study compares them with current methodologies. This study selects the most informative characteristics from liver MRI medical image datasets using Genetic and spider algorithm. A thorough comparative analysis is carried out to assess the algorithm's performance in comparison to conventional feature selection techniques as well as the other techniques of nature-inspired methods. To ensure compatibility with the chosen Hybrid Genetic and Social Spider [HGSS] algorithms preparing hepatic medical imaging datasets is part of the experimental procedure. Genetic algorithm is used to extract the best features from the MR images and the nature-inspired social spider optimization algorithm is used to optimize the selected features. Evaluation metrics are used to measure how well feature selection performs in terms of computing efficiency Accuracy, Precision, Recall and F-Score. The proposed hybrid approach was compared with the traditional Random Forest (RF) model for image classification. The proposed HGSS algorithm optimization of the feature selection with CNN layer with 32 filter size of 3 X 3 showed the performance better than the RF.