Deep learning-based model for detecting 2019 novel coronavirus pneumonia on high-resolution computed tomography

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Computed tomography (CT) is crucial for diagnosing COVID-19 pneumonia.
• Deep learning offers potential for automated detection of COVID-19 pneumonia in CT scans.

Purpose of the Study:

• To develop and validate a deep learning system for detecting COVID-19 pneumonia on high-resolution CT.
• To assess the system's performance against expert radiologists and evaluate its impact on efficiency.

Main Methods:

• Retrospective collection of 46,096 anonymous CT images from 106 patients (51 COVID-19 positive, 55 controls).
• Prospective evaluation on 27 patients to compare model performance with radiologists.
• External validation conducted at a separate hospital to assess robustness.

Main Results:

• The deep learning model achieved high accuracy: 95.24% per patient and 98.85% per image in the internal dataset.
• Performance was comparable to expert radiologists in prospective evaluation.
• External dataset accuracy reached 96%, demonstrating robustness.
• Radiologist reading time decreased by 65% with AI assistance.

Conclusions:

• The developed deep learning model demonstrates high accuracy and robustness in detecting COVID-19 pneumonia on CT scans.
• The system performs comparably to expert radiologists.
• AI assistance significantly enhances radiologist efficiency in clinical practice.