Related Experiment Video
Updated: Sep 12, 2025
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
Improving the accuracy of emergency department clinicians in detecting SARS-COV-2 on chest X-rays using a bespoke
Jasdeep Bahra1,2, Anita Acharya1, Sarim Ather3
1Emergency Medicine Research Oxford (EMROx), Oxford University Hospitals NHS Foundation Trust, Headley Way, Oxford, OX3 9DU, UK.
Background:
During and after the COVID pandemic, online learning became a key component in most undergraduate and post-graduate training. The non-specific symptoms of SARS-CoV-2 and limitations of available diagnostic tests can make it difficult to detect and diagnose in acute care settings. Accurate identification of SARS-CoV-2 related changes on chest x-ray (CXR) by frontline clinicians involved in direct patient care in the Emergency Department (ED) is an important skill. We set out to measure the accuracy of ED clinicians in detecting SARS-CoV-2 changes on CXRs and assess whether this could be improved using an online learning platform.
Methods:
Baseline reporting performance of a multi-centre cohort of ED clinicians with varying experience was assessed via the Report and Image Quality Control (RAIQC) online platform. Emergency Medicine clinicians working in EDs across five hospitals in the Thames Valley Emergency medicine Research Network (TaVERN) region were recruited over a six-month period. An image bank was created containing both SARS-CoV-2 and non- SARS-CoV-2 pathological findings. Radiological ground truth diagnosis was established by thoracic radiologists and corroborated by RT- PCR results. Participants then undertook an online training module with performance re-assessed. Diagnostic accuracy and speed of X-ray reporting was assessed before and after training in 3 subgroups: Consultants, Junior Doctors and Nurses.
Results:
90 clinicians undertook pre-training assessment and 56 undertook post training assessment. There was an overall improved reporting accuracy for participants who undertook both pre and post training assessments from (44.0±10.5%) to 57.4% (±9.39)% (p < 0.001). The sensitivity for recognition of SARS-CoV-2 improved from 64.8 to 76.8%.
Conclusion:
ED clinicians show moderate baseline accuracy in the identification of SARS-CoV-2 related changes on CXR. Accuracy and speed can be improved by online training.
