Conv-CapsNet: capsule based network for COVID-19 detection through X-Ray scans.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• The rapid global spread of Coronavirus (COVID-19) necessitated efficient detection methods.
• Radiological images like X-rays and CT scans are crucial for identifying COVID-19 infections.
• Deep learning models show promise in analyzing medical images for disease detection.

Purpose of the Study:

• To propose a shallow architecture using Capsule Networks with convolutional layers for COVID-19 detection from X-ray images.
• To develop a fast, robust, and efficient model for classifying X-ray images into COVID-19, No Findings, and Viral Pneumonia categories.
• To evaluate the model's performance with limited training data.

Main Methods:

• A shallow Capsule Network architecture combined with convolutional layers was designed for feature extraction and spatial information understanding.
• The model was trained on X-ray datasets to classify images into three distinct categories.
• A 5-fold cross-validation strategy was employed to assess the model's generalization capabilities.

Main Results:

• The proposed shallow model, with 23 million parameters, demonstrated strong performance despite requiring fewer training samples.
• Achieved an average accuracy of 96.47% for multi-class classification (COVID-19, No Findings, Viral Pneumonia).
• Attained an average accuracy of 97.69% for binary classification (COVID-19 vs. others) on the X-ray dataset.

Conclusions:

• The developed Capsule Network-based model offers an effective and efficient solution for COVID-19 detection using X-ray imaging.
• The model's shallow architecture and high accuracy make it a valuable tool for researchers and medical professionals.
• This approach can assist in the diagnosis and prognosis of COVID-19 patients, contributing to pandemic control efforts.