[Paper Review] Rapid AI Development Cycle for the Coronavirus (COVID-19) Pandemic: Initial Results for Automated Detection & Patient Monitoring using Deep Learning CT Image Analysis

Ophir Gozes, Maayan Frid-Adar+6 more|arXiv (Cornell University)|Mar 10, 2020

COVID-19 diagnosis using AI9 references672 citations

TL;DR

This study presents a rapidly developed deep learning system for automated detection and monitoring of COVID-19 using non-contrast thoracic CT scans. The model achieves 0.996 AUC (98.2% sensitivity, 92.2% specificity) in distinguishing COVID-19 from non-COVID cases and provides quantitative opacity measurements and a dynamic 'Corona score' for tracking disease progression over time.

ABSTRACT

Purpose: Develop AI-based automated CT image analysis tools for detection, quantification, and tracking of Coronavirus; demonstrate they can differentiate coronavirus patients from non-patients. Materials and Methods: Multiple international datasets, including from Chinese disease-infected areas were included. We present a system that utilizes robust 2D and 3D deep learning models, modifying and adapting existing AI models and combining them with clinical understanding. We conducted multiple retrospective experiments to analyze the performance of the system in the detection of suspected COVID-19 thoracic CT features and to evaluate evolution of the disease in each patient over time using a 3D volume review, generating a Corona score. The study includes a testing set of 157 international patients (China and U.S). Results: Classification results for Coronavirus vs Non-coronavirus cases per thoracic CT studies were 0.996 AUC (95%CI: 0.989-1.00) ; on datasets of Chinese control and infected patients. Possible working point: 98.2% sensitivity, 92.2% specificity. For time analysis of Coronavirus patients, the system output enables quantitative measurements for smaller opacities (volume, diameter) and visualization of the larger opacities in a slice-based heat map or a 3D volume display. Our suggested Corona score measures the progression of disease over time. Conclusion: This initial study, which is currently being expanded to a larger population, demonstrated that rapidly developed AI-based image analysis can achieve high accuracy in detection of Coronavirus as well as quantification and tracking of disease burden.

Motivation & Objective

• To develop an AI-based system for automated detection of COVID-19 in thoracic CT scans.
• To enable quantitative monitoring of disease burden and progression in infected patients.
• To support radiologists and clinicians in triaging high-volume imaging during the pandemic.
• To evaluate the performance of the system on international datasets from China and the U.S.

Proposed method

• The system employs a combination of 2D and 3D deep learning models adapted from existing AI platforms.
• It integrates clinical knowledge of COVID-19 CT features, such as ground-glass opacities and consolidative patterns.
• The model performs slice-based heat map visualization and 3D volume rendering of lung opacities.
• A novel 'Corona score' is computed to quantitatively track disease burden across time points.
• The system was trained and validated on a retrospective dataset of 157 international patients, including cases from China and the U.S.
• Performance was evaluated using AUC, sensitivity, and specificity on a test set of confirmed and control cases.

Experimental results

Research questions

• RQ1 Can a deep learning system accurately differentiate between COVID-19 and non-COVID-19 cases using thoracic CT scans?
• RQ2 Can the system provide reliable, quantitative measurements of lung opacity volume and distribution in infected patients?
• RQ3 How well can the system track changes in disease burden over time using serial CT scans?
• RQ4 Can the proposed 'Corona score' serve as a consistent, objective metric for monitoring disease progression or resolution?
• RQ5 Does the system maintain high performance across diverse international patient populations?

Key findings

• The system achieved an area under the curve (AUC) of 0.996 (95% CI: 0.989–1.00) in distinguishing COVID-19 from non-COVID-19 cases on a test set of Chinese patients.
• At a high-sensitivity working point, the system demonstrated 98.2% sensitivity and 92.2% specificity.
• For patients with confirmed COVID-19, the system provided accurate volumetric measurements of opacities and visualized them via heat maps and 3D reconstructions.
• The 'Corona score' successfully tracked disease progression and resolution, showing a 49% reduction in opacity burden over 4 days in one patient.
• In a longitudinal case study, the system detected complete resolution of opacities (Corona score: 0) after 15 days, indicating full recovery.
• The system demonstrated consistent performance across international datasets, including patients from China and the U.S.