Machine Learning Models for Development and Evaluation of Foldscope and Microscopy as Malaria Diagnostic Tool Using Invasive Samples

This study evaluates the development and effectiveness of a Foldscope-based diagnostic tool for detecting malaria parasites, comparing its performance to conventional microscopy. Invasive (blood) samples were collected from consented 50 asymptomatic and 50 symptomatic patients who have provided consent. Additionally, interview were conducted with 20 medical personnel. This was done with approval from University of Medical Sciences Teaching Hospital (UNIMEDTH) in Akure, Ondo State, Nigeria. The cross-sectional study design aimed to assess diagnostic accuracy using metrics such as sensitivity, specificity, and overall accuracy. Chi-square Test showed no statistically significant relationships between age and gender groups and the outcome (p > 0.05). Malaria parasite density was below the normal threshold (2483.20 ±437.05 parasite per microliter (parasites/μL)), while parasitized RBCs exceeded normal levels (49.66%). Hemoglobin (11.67 ±2.47 g/dL) and packed cell volume (35.08 ±7.27%) were lower than normal, suggesting potential anemia in infected individuals. Microscopy exhibited higher sensitivity (67%), specificity (75%), and accuracy (70%) compared to Foldscope, which had sensitivity of 56%, specificity of 67%, and accuracy of 60%, making Microscopy the more effective diagnostic tool. Foldscope demonstrates better performance than Microscopy in F1 score (82% vs. 62%) and AUC (0.667 vs. 0.65), indicating better overall diagnostic performance, except in recall, where both methods performed equally well. Medical personnel reported Foldscope easy to use (70%), with 80% satisfied with image quality and 80% found it adequate. Most (84%) recommended it for malaria diagnosis, while only a small percentage disagreed. Overall results indicated that microscopy surpasses the Foldscope in sensitivity and accuracy, but the Foldscope demonstrated potential as a low cost-effective and portable alternative for malaria diagnosis, particularly in resource-limited settings.