COVID Detection and Severity Prediction with 3D-ConvNeXt and Custom Pretrainings
This work addresses COVID-19 diagnosis and monitoring for medical applications, but it is incremental as it adapts an existing model to a new data type.
The paper tackles the problem of detecting COVID-19 infection and predicting lung damage severity from 3D CT scans by adapting the ConvNeXt model to process three-dimensional data and designing custom pretraining methods, achieving 2nd place in a severity detection challenge and 3rd in a detection challenge.
Since COVID strongly affects the respiratory system, lung CT-scans can be used for the analysis of a patients health. We introduce a neural network for the prediction of the severity of lung damage and the detection of a COVID-infection using three-dimensional CT-data. Therefore, we adapt the recent ConvNeXt model to process three-dimensional data. Furthermore, we design and analyze different pretraining methods specifically designed to improve the models ability to handle three-dimensional CT-data. We rank 2nd in the 1st COVID19 Severity Detection Challenge and 3rd in the 2nd COVID19 Detection Challenge.