Analyzing Malaria Disease Using Effective Deep Learning Approach.

Area of Science:

• Medical Imaging
• Computational Biology
• Parasitology

Background:

• Medical specialists utilize image processing and computer-aided diagnostic systems for malaria treatment decisions.
• Accurate malaria detection from blood smear images is crucial for patient monitoring, though atypical cases present challenges.

Purpose of the Study:

• To evaluate the effectiveness of various deep learning models for malaria detection using thin blood smear images.
• To enhance the performance of computer-aided diagnostic systems for optimal malaria detection.

Main Methods:

• Analysis of 7000 malaria images using convolutional neural network (CNN) models: Xception, Inception-V3, ResNet-50, NasNetMobile, VGG-16, and AlexNet.
• Implementation of a rotational method to improve training and validation datasets.
• Utilization of the Mish activation function and Nadam optimizer with the Xception model.

Main Results:

• The Xception model, with Mish activation and Nadam optimizer, demonstrated superior performance in classifying malaria from blood smear images.
• A combined score of 99.28% was achieved for performance metrics including recall, accuracy, precision, and F1 measure.
• An accuracy level of 98.86% was obtained for evaluating images outside the primary dataset.

Conclusions:

• Deep learning models, particularly Xception, offer a highly effective approach for accurate malaria detection from medical images.
• The developed computer-aided diagnostic system shows significant potential for improving malaria diagnosis and patient management.