“Multi-modality AI for longitudinal outcome assessment of COVID-19”
Advance-CTR
As of March 2023, there were more than 100 million confirmed cases of COVID-19 in the US. As such, efficient, accurate triage is critical to alleviate the burden of healthcare systems and improve patient outcomes. Early prediction of disease severity could assist in timely resource allocation. Medical recording and imaging such as chest X-rays and vital signals are vital for prognosis, especially in rural settings. Artificial intelligence (AI) has been an emerging technology in the field of digital healthcare that has been applied to COVID-19. Compared to the traditional workflow with only human readers, AI enables safer, more accurate, and more efficient image analysis. Our team has developed AI models for the diagnosis and progression of COVID-19. Particularly, we have demonstrated that AI models based on the combination of imaging and clinical data perform promisingly in diagnosis and prognosis. However, the heterogeneity of multi-modality data in large-scale cohort settings usually leads to decreased performance. Besides, the long COVID illness further prioritizes the longitudinal assessment, which helps to improve the management of high-risk patients. In this talk, I will present our COVID-19 AI studies for prognosis and longitudinal assessment of patients’ outcomes.
This study is sponsored through Advance-CTR supported by the IDeA-CTR grant (U54GM115677)