Exploring the Impact of Batch Size on Deep Learning Artificial Intelligence Models for Malaria Detection.

Area of Science:

• Medical diagnostics
• Artificial intelligence in healthcare
• Parasitology

Background:

• Malaria diagnosis relies on microscopy, which is time-consuming and labor-intensive.
• Artificial intelligence (AI) offers automated screening of blood smears, overcoming limitations of traditional methods.
• Convolutional neural networks (CNNs) are a type of AI being explored for medical image analysis.

Purpose of the Study:

• To investigate the impact of batch size on the accuracy and training speed of CNN models for malaria detection.
• To evaluate how different batch sizes affect the performance of AI models in analyzing thin blood smear images.

Main Methods:

• Utilized the NIH-NLM-ThinBloodSmearsPf dataset containing Plasmodium falciparum images.
• Developed four identical 10-layer CNN models trained with varying batch sizes (16 to 128) for 10 epochs.
• Collected model prediction accuracy, training time, and F1-scores for 16 different training configurations.

Main Results:

• All CNN models achieved high F1-scores (94%-96%).
• Increasing batch size from 16 to 128 led to a 1% decrease in average F1-score accuracy.
• Larger batch sizes reduced average training time by 28.11%.

Conclusions:

• Smaller batch sizes demonstrated improved accuracy in CNN-based malaria detection.
• Optimizing batch size is crucial for developing clinically viable AI diagnostic tools.
• Smaller batch sizes may enable more accessible and cost-effective AI model training for malaria screening.