Hybrid Deep Neural Network for Classification of Malaria Parasites in Microscopic Images

Malaria is a critical epidemic disease transmitted by bites from female Anopheles mosquitoes, caused by Plasmodium parasites. Although it does not spread directly between people, early detection is essential to prevent mild cases from progressing. This research develops a hybrid deep-learning model for classifying malaria parasites in thin blood smears including infected and uninfected erythrocytes. The model employs a convolutional neural network (CNN) with recurrent neural network (RNN) variants. To advance interpretability, the Explainable AI technique, Gradient-weighted Class Activation Map (GradCAM), is incorporated to highlight key regions in the image that support the model's decision-making. Additionally, regularization techniques, including data augmentation, batch normalization, and L2 regularization, are applied to enhance model performance. The proposed CNN-BiGRU model achieved a validation accuracy of 96.7%, precision of 0.97%, recall of 0.96%, F1-score of 0.97%, and an AUC of 0.99 on both training and testing sets.