LCCNN: a Lightweight Customized CNN-Based Distance Education App for COVID-19 Recognition.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Distance Education

Background:

• The global epidemic highlighted the importance of distance learning in education.
• Medical diagnosis and education require accessible and effective tools, especially during health crises.
• Convolutional Neural Networks (CNNs) show promise in medical image analysis.

Purpose of the Study:

• To develop a web-based application for COVID-19 recognition using a novel CNN model.
• To evaluate the model's accuracy and diagnostic capabilities.
• To create an explainable AI tool for medical education and patient assessment.

Main Methods:

• Proposed an 8-layered lightweight, customized convolutional neural network (LCCNN).
• Implemented a five-channel data augmentation technique to prevent overfitting.
• Developed a user-friendly web-based application integrating the LCCNN model.

Main Results:

• The LCCNN model achieved an accuracy of 91.78%, outperforming eight other state-of-the-art methods.
• The model successfully detected COVID-19 lesions in CT images.
• Generated heat maps provided clear visualization of lesions for explainability.

Conclusions:

• The developed web-based app and LCCNN offer a valuable diagnostic perspective for COVID-19 patients.
• The tool serves as an effective platform for distance medical education, enhancing student interaction and understanding.
• The explainable nature of the LCCNN model benefits both radiologists and educators.