XMal-CNN: An Explainable Deep Neural Model for Automated Malaria Detection from Blood Smear Images

Malaria is a severe and critical health issue widespread throughout the globe. Malaria must be diagnosed correctly and efficiently in its initial stage in order to treat and cure it before it becomes a terminal illness. The current paper describes XMal-CNN, a novel deep learning approach to be utilized in automated malaria diagnosis from microscopic blood smear images. The proposed structure utilizes a depth-wise Convolutional Neural Network (CNN) with a Squeeze and Excitation (SE) block to increase feature representation and perform classification of images. The suggested approach model performs in such a way that it surpasses baseline CNNs and currently existing state-of-the-art approaches, achieving 95.26% accuracy, 93.97% precision, 96.73% recall, and 95.33% F1-score. To improve model interpretability and explainability, Explainable AI (XAI) techniques such as LIME and Grad-CAM++ are used, providing useful insights and understanding of the decision making process of the model. Systematic and extensive evaluations on benchmark blood smear image datasets are conducted to validate the performance and explainability of the proposed model. Due to its superior diagnostic precision and interpretability, XMal-CNN becomes a trustworthy and important AI-assisted tool, aiding healthcare experts in making informed and data-driven decisions to diagnose and treat malaria.