Deep learning assisted COVID-19 detection using full CT-scans.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• The COVID-19 pandemic highlighted the need for efficient and accurate diagnostic tools.
• Current medical institutions face limitations in handling large-scale diagnostic demands.
• Automated diagnosis systems are crucial for improving speed, accuracy, and accessibility in healthcare.

Purpose of the Study:

• To propose and evaluate an automated deep learning model for COVID-19 detection using CT scans.
• To develop a computationally efficient model suitable for deployment on edge devices.
• To achieve high accuracy in identifying COVID-19 infection from chest CT images.

Main Methods:

• Utilized a deep learning technique based on a residual network (ReCOV-101) with skip connections.
• Preprocessed chest CT scans using segmentation and interpolation to enhance detection accuracy.
• Trained the model on a single enterprise-level GPU for reduced computational requirements.

Main Results:

• The ReCOV-101 model achieved an accuracy of 94.9% in detecting COVID-19 infection.
• The model demonstrated excellent performance with reduced hardware intensity.
• The approach allows for potential integration with medical equipment for streamlined examination.

Conclusions:

• The proposed ReCOV-101 model offers an effective and efficient automated solution for COVID-19 diagnosis from CT scans.
• The model's low hardware requirements facilitate edge deployment, reducing cloud dependency.
• This research contributes to advancing medical imaging analysis and diagnostic capabilities in pandemic scenarios.