eduzhai > Applied Sciences > Engineering >

Was there COVID-19 back in 2012? Challenge for AI in Diagnosis with Similar Indications

  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Purpose: Since the recent COVID-19 outbreak, there has been an avalanche ofresearch papers applying deep learning based image processing to chestradiographs for detection of the disease. To test the performance of the twotop models for CXR COVID-19 diagnosis on external datasets to assess modelgeneralizability. Methods: In this paper, we present our argument regarding theefficiency and applicability of existing deep learning models for COVID-19diagnosis. We provide results from two popular models - COVID-Net and CoroNetevaluated on three publicly available datasets and an additional institutionaldataset collected from EMORY Hospital between January and May 2020, containingpatients tested for COVID-19 infection using RT-PCR. Results: There is a largefalse positive rate (FPR) for COVID-Net on both ChexPert (55.3 ) and MIMIC-CXR(23.4 ) dataset. On the EMORY Dataset, COVID-Net has 61.4 sensitivity, 0.54F1-score and 0.49 precision value. The FPR of the CoroNet model issignificantly lower across all the datasets as compared to COVID-Net -EMORY(9.1 ), ChexPert (1.3 ), ChestX-ray14 (0.02 ), MIMIC-CXR (0.06 ).Conclusion: The models reported good to excellent performance on their internaldatasets, however we observed from our testing that their performancedramatically worsened on external data. This is likely from several causesincluding overfitting models due to lack of appropriate control patients andground truth labels. The fourth institutional dataset was labeled using RT-PCR,which could be positive without radiographic findings and vice versa.Therefore, a fusion model of both clinical and radiographic data may havebetter performance and generalization.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...