Novel Light Convolutional Neural Network for COVID Detection with Watershed Based Region Growing Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• The COVID-19 pandemic necessitated rapid and accurate diagnostic tools.
• Medical imaging, including CT scans and X-rays, plays a vital role in COVID-19 detection.
• Artificial intelligence (AI) offers potential for automated and precise diagnostic systems.

Purpose of the Study:

• To develop a novel, lightweight Convolutional Neural Network (CNN) model for automated COVID-19 detection.
• To improve the accuracy and efficiency of diagnosing COVID-19 from lung imaging data.
• To evaluate the model's performance on both Chest X-ray (CXR) and Computed Tomography (CT) scans.

Main Methods:

• Development of a novel, lightweight CNN architecture.
• Integration of watershed-based region-growing segmentation for image analysis.
• Utilized a dataset comprising both Chest X-ray (CXR) and Computed Tomography (CT) scans.
• Employed 5-fold cross-validation for robust performance evaluation.

Main Results:

• The proposed lightweight CNN model achieved high diagnostic accuracy.
• Achieved a mean accuracy of 98.8% on X-ray images and 98.6% on CT scans.
• Demonstrated high Positive Predictive Value (PPV) of 0.99% and Negative Predictive Value (NPV) of 0.98% on X-rays, and 0.97% PPV and 0.99% NPV on CT scans.

Conclusions:

• The developed lightweight CNN model offers a promising approach for rapid and accurate COVID-19 diagnosis.
• The model outperforms previous state-of-the-art methods in detecting COVID-19 from medical images.
• This AI-driven tool can aid clinicians in timely diagnosis and patient management during epidemics.