MicrobeNet: An Automated Approach for Microbe Organisms Prediction Using Feature Fusion and Weighted CNN Model

Microbial organisms are everywhere, millions residing within the human body and also cover 60% of the living earth. These microbes can pose significant health risks, causing diseases such as malaria and toxoplasmosis. Toxoplasmosis is notably prevalent, with seroprevalence rates ranging from 3.6 to 84% in an African region, underscoring the necessity for automated microorganism detection techniques. This research work aims to predict the presence of microorganisms in the human body. We propose a novel approach that combines and integrates principal-component analysis, Chi-square, and analysis-of-variance features using a weighted convolutional-neural-network model called MicrobeNet. The results highlight the efficacy of the proposed method, achieving a remarkable 99.97% in accuracy, recall, precision, and F1-score. The experiments use multiple deep and machine learning models to detect ten distinct microbial forms. The results of the proposed model are compared with those of previously published research. Additionally, k-fold cross validation confirms the robustness of these findings. This research significantly advances the field of microbiology by providing a highly accurate method for microorganism identification, facilitating early disease detection and prevention.