Malaria detection through digital microscopic imaging using Deep Greedy Network with transfer learning.

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Healthcare

Background:

• Manual microscopy for malaria diagnosis is labor-intensive and prone to errors.
• Automated methods using deep learning show promise but struggle with small datasets.
• Convolutional Neural Networks (CNNs) are being explored for malaria parasite detection.

Purpose of the Study:

• To develop a comprehensive computer-aided diagnosis scheme for automated malaria parasite detection in thin blood smear images.
• To optimize feature selection using transfer learning and deep CNNs.
• To enhance prediction accuracy for Plasmodium falciparum detection.

Main Methods:

• Proposed a deep CNN scheme integrating transfer learning for feature selection.
• Utilized ResNet 152 model combined with Deep Greedy Network for training.
• Extracted layer embeddings from intermediate convolutional layers for feature cross-checking.

Main Results:

• Evaluated the hybrid model using accuracy, precision, recall, specificity, and F1-score.
• Compared the proposed model's performance against existing deep learning algorithms.
• Demonstrated promising outcomes and enhanced prediction quality.

Conclusions:

• The hybrid deep CNN model shows superior performance compared to other models based on accuracy metrics.
• Layer embedding extraction allows for manual verification of the model's performance.
• The developed scheme offers a reliable and automated approach for malaria diagnosis.