ENHANCING BIOMEDICAL IMAGE CLASSIFICATION: EXPLORING VARIOUS CLASSIFIERS FOR ACCURATE ANALYSIS
Jaya Bijaya Arjun Das, Debasmita Behera · 2023
The primary focus of this study revolves around enhancing disease diagnosis, analysis, and detection through optimized models, particularly for biomedical images. In the medical field, professionals heavily rely on pathological reports for diagnosis. However, visualizing these reports can greatly aid doctors in their diagnostic process. This research specifically aims to develop a streamlined model for swift analysis and diagnosis. The study encompasses data related to Brain Tumors, Skin Cancer, Liver Disorders, and Breast Cancer. The methodology involves several steps: starting from raw data, utilizing Convolutional Neural Networks (CNNs) for image data, extracting features from the data, and optimizing these features for machine learning models. The machine learning algorithms employed in this research include the k-nearest neighbors' algorithm (KNN), Random Forest (RF), and Support Vector Machine (SVM). These algorithms are evaluated across all cancer types under consideration. Distinct accuracies are observed from different machine learning classifiers. Notably, CNN, unlike other models, does not rely on external features due to its intrinsic capability for feature extraction, selection, and integration. On the other hand, for the other machine learning techniques, feature acquisition is crucial for the classifier model. In certain cases of cancer data, Support Vector Machine (SVM) demonstrates higher accuracy, while CNN performs best when external features are not utilized. Interestingly, through the process of Brain Storm Optimization (BSO), optimizing features and applying them to the SVM model yields exceptional results compared to alternative methods.