Generalized fractional optimization-based explainable lightweight CNN model for malaria disease classification.

Area of Science:

• Artificial Intelligence
• Medical Diagnostics
• Computational Biology

Background:

• Deep learning (DL) excels in healthcare image processing, offering advantages over traditional methods.
• Malaria, caused by Plasmodium falciparum, poses a significant global public health threat.
• Existing DL models for malaria diagnosis show promise but face challenges in computational efficiency and interpretability.

Purpose of the Study:

• To develop a generalized fractional order-based explainable lightweight convolutional neural network (CNN) for malaria diagnosis.
• To address the limitations of computational inefficiency and lack of interpretability in current DL approaches.
• To provide a cost-effective and time-efficient diagnostic tool for malaria.

Main Methods:

• Proposed a novel lightweight CNN architecture incorporating fractional order optimization.
• Trained and validated the model using the standard NIH dataset, the MP-IDB dataset, and the M5 test set.
• Evaluated model performance using accuracy, precision, recall, and F1-score metrics.

Main Results:

• Achieved 95% accuracy on the NIH dataset, outperforming complex existing models in speed and effectiveness.
• Demonstrated robust generalizability with 92% accuracy on the MP-IDB dataset and 90.4% on the M5 test set.
• The model's efficacy was further confirmed by strong precision, recall, and F1-score values.

Conclusions:

• The proposed fractional order-based explainable lightweight CNN offers an improved and efficient solution for malaria diagnosis.
• The model's high accuracy, speed, and generalizability highlight its potential for real-world public health applications.
• This research contributes to advancing AI-driven diagnostic tools for infectious diseases.