- Objective:
Lung Ultrasound (LUS) imaging has emerged as an effective bedside tool for monitoring COVID-19 patients. Numerous AI-based applications have been developed to aid in the diagnosis and identification of COVID-19 lung biomarkers. Recently, various AI-based techniques have been proposed to score patient severity, utilizing video- and frame-based labels. However, manually labeling each individual frame in an ultrasound video clip is a time-consuming and expensive process. It would be more practical to label the entire ultrasound video clip and treat the severity label of the video clip as a substitute for the corresponding frames. However, employing this method introduces inaccuracies in labeling since not all frames in a clip demonstrate the same level of severity.
In this project, we use the same 4-level ultrasound scoring scheme as defined in Paola et al, 2022 (image below) and will proposed a AI-based approach for the prediction of ultrasound video-level severity score and ultimately extent to accurate frame-level severity score prediction.
