A CNN-Based Malaria Diagnosis from Blood Cell Images with SHAP and LIME Explainability

High diagnostic accuracy could improve malaria detection in resource-limited settings: The custom CNN achieved 96% accuracy with precision and recall scores exceeding 0.95, suggesting that AI-based blood cell image analysis could provide reliable malaria diagnosis in remote areas where expert microscopy is unavailable or impractical.

Explainability tools make AI decisions transparent and trustworthy for medical use: The paper demonstrates that SHAP, LIME, and Saliency Maps can reveal which features of blood cell images the AI model uses to make diagnostic decisions, helping clinicians understand and verify that the system is identifying medically relevant patterns rather than learning spurious correlations.

Multiple AI architectures were tested to ensure robust performance: By comparing the custom CNN against established models (ResNet50, VGG16, MobileNetV2, and DenseNet121), the study provides evidence that the approach is not dependent on a single model design, which strengthens confidence in the method's reliability for real-world deployment.

AI-based diagnosis addresses real limitations of current methods: Traditional microscopic blood smear analysis suffers from low sensitivity, expert dependency, and resource constraints—problems that automated image analysis could meaningfully reduce, particularly benefiting populations in tropical and subtropical regions with high malaria burden.