MalNet-DAF: Dual-Attentive Fusion Deep Learning Model for Malaria Parasite Classification

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Parasitology

Background:

• Malaria diagnosis is critical for treatment but conventional methods face challenges like human error and data limitations.
• Accurate and rapid malaria detection is essential to combat Plasmodium parasite infections.

Purpose of the Study:

• To introduce MalNet-DAF, a novel deep learning model for enhanced malaria-infected cell diagnosis and classification.
• To improve the reliability and accuracy of malaria diagnostics through advanced computational techniques.

Main Methods:

• Developed MalNet-DAF, integrating Convolutional Neural Networks (CNNs) for spatial features and Bidirectional Long Short-Term Memory (Bi-LSTM) for temporal dependencies.
• Incorporated Spatial Attention Module (SAM) and Temporal Attention Module (TAM) for feature refinement and informative time step selection.
• Trained and validated the model on a National Institutes of Health (NIH) malaria dataset.

Main Results:

• MalNet-DAF achieved a high classification accuracy of 99.24%.
• The model outperformed traditional baseline diagnostic methods.
• Dynamic attention mechanisms enhanced model interpretability and performance.

Conclusions:

• Dynamic attention-driven deep learning shows significant potential for real-time clinical decision-making in malaria diagnosis.
• MalNet-DAF offers a promising solution to overcome limitations in current healthcare diagnostics for infectious diseases.