Machine Learning-Based Automated Hemoparasite Detection (MLAHD) Model: A Mathematical Perspective

Hemoparasitic infections, particularly those caused by Plasmodium spp., pose significant health challenges worldwide, especially in resource-limited settings where traditional diagnostic methods are often inadequate. This study addresses critical gaps in existing diagnostic practices, including the reliance on labour-intensive microscopy and the lack of accessible, automated solutions. The aim is to develop and evaluate an innovative Machine Learning-Based Automated Hemoparasite Detection (MLAHD) model that integrates Convolutional Neural Networks (CNNs) with affordable Raspberry Pi systems to enhance diagnostic accuracy and accessibility. This methodology is grounded in a robust mathematical framework that facilitates image preprocessing, feature extraction, and classification. A diverse dataset of 10,000 labelled blood smear images, annotated by expert pathologists, was utilized to train the model. Key techniques applied included normalization, segmentation, and hierarchical feature extraction using CNN architectures, alongside the integration of explainable AI to enhance interpretability. Results from pilot studies conducted in resource-limited settings demonstrated that the MLAHD model achieved high sensitivity (over 90%) and specificity, significantly outperforming conventional diagnostic methods. Users’ feedback highlighted the system's ease of usage and rapid diagnostic capabilities, underscoring its potential to transform hemoparasite detection in underserved communities. This research contributes to advancements in biomedical engineering by providing a scalable, cost-effective solution that improves diagnostics for hemoparasitic diseases, underpinned by a solid mathematical foundation.