- We present a deep learning solution on chest radiographs (CXRs) for COVID-19 diagnosis, PROS CXR: COVID-19, in close collaborations with KAIST (Korea Advanced Institute of Science and Technology).
- Our research on COVID-19 diagnosis on CXRs was published on a renowned peer-reviewed journal, IEEE Transactions on Medical Imaging, in August 2020.
- Our solution classify CXRs into 4 classes with high accuracy: Normal, bacterial pneumonia, tuberculosis, and viral pneumonia including COVID-19.
- Our solution achieves state-of-the-art performance (accuracy: 92.66% and F1-score: 92.71%) and provides clinically interpretable saliency maps, which are useful for COVID-19 diagnosis and patient triage.
- Provides clinician-friendly user interface with features of LIST, DICOM TAGS, VIEW, RESULTS, TOOLS, and DOWNLOAD.
- On-Premise standalone or cloud-based solutions are available with native DICOM/DICOMweb protocol supports.
- Real-time anonymization, TLS-encrypted secured connection, and image sharing functionality in our cloud-based solution.