Designing the CORI score for COVID-19 diagnosis in parallel with deep learning-based imaging models.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Infectious Disease Diagnostics

Background:

• The COVID-19 pandemic strained global healthcare, highlighting the need for accessible diagnostic alternatives to RT-PCR, particularly in resource-limited settings.
• Chest X-rays combined with AI offer a potential solution for rapid and widespread COVID-19 detection.

Purpose of the Study:

• To develop an AI-assisted diagnostic model integrating chest X-ray images and clinical data.
• To create a COVID-19 Risk Index (CORI) Score using a deep learning ResNet architecture.
• To evaluate the diagnostic performance of the AI model and CORI Score.

Main Methods:

• A multicenter cohort study in Jakarta, Indonesia, involving 367 participants (COVID-19 positive, non-COVID-19 pneumonia, healthy controls).
• Collection of chest X-ray images, clinical parameters (fever, cough, oxygen saturation), and laboratory findings (D-dimer, C-reactive protein).
• Development and validation of a ResNet-based deep learning model and the CORI Score using integrated data.

Main Results:

• The ResNet model achieved 91% accuracy, 94% sensitivity, and 92% specificity in internal validation, and identified 82% of COVID-19 cases in external validation.
• The integrated approach (imaging, clinical, lab data) yielded an AUC of 0.98 and sensitivity >95%.
• The CORI Score demonstrated high diagnostic performance: 96.6% accuracy, 98% sensitivity, 95.4% specificity, 99.5% NPV, and 91.1% PPV.

Conclusions:

• The ResNet-based AI model and CORI Score show significant potential as diagnostic tools for COVID-19.
• The AI model's performance is comparable to experienced thoracic radiologists in Indonesia.
• These AI-driven tools could improve COVID-19 diagnosis, especially in resource-constrained environments.