Computer-aided diagnosis of Long COVID from lung X-Ray images
Research Focus
Despite the end of the COVID-19 pandemic, chronic health effects from the virus may persist for months following infection in a condition known as Long COVID. However, there is currently a lack of specific and robust criteria for Long COVID diagnosis. Given the potential for medical imaging to be used in diagnosis, we identify a vision transformer-based method to specifically detect Long COVID from X-ray images.
Key Findings
After data pre-processing and hyperparameter optimization, our ViT-Base16 model achieved a 0.96 F1-score across Long COVID, Ongoing COVID, and Normal classes. This model may help facilitate more accurate and efficient diagnoses of Long COVID compared to existing self-reporting methods.