Detection of Pneumonia from Chest X-ray Images Utilizing MobileNet Model.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Pneumonia is a leading global cause of death, often visually similar to other respiratory diseases like tuberculosis.
• Variability in chest X-ray acquisition and processing complicates accurate pneumonia detection.
• Developing robust, data-driven algorithms for pneumonia identification is crucial.

Purpose of the Study:

• To develop and validate a deep-learning model for distinguishing between normal and severe pneumonia cases using chest X-ray images.
• To compare the performance of eight pre-trained deep learning models for pneumonia classification.

Main Methods:

• A deep-learning system was designed, incorporating eight pre-trained models: ResNet50, ResNet152V2, DenseNet121, DenseNet201, Xception, VGG16, EfficientNet, and MobileNet.
• These models were trained and evaluated on two distinct chest X-ray datasets (5,856 and 112,120 images).
• Hyperparameter tuning, including batch sizes, epochs, and optimizers, was performed for model optimization.

Main Results:

• The MobileNet model demonstrated the highest classification accuracy, achieving 94.23% on one dataset and 93.75% on the other.
• Comparative analysis identified MobileNet as the most effective model among the eight tested.
• The study systematically evaluated model performance based on key hyperparameters.

Conclusions:

• Deep learning models, particularly MobileNet, show significant promise for accurate pneumonia detection from chest X-rays.
• The findings highlight the potential of AI in improving the diagnosis of pneumonia, addressing challenges posed by image variability.
• Further validation using diverse imaging techniques and expert analysis is recommended.