Multi-Centre Optimization and Validation of an Open Deep Learning Model for Covid-19 Detection on Chest Radiographs
| Main Authors: | Kim-Ann Git, Aida binti Abdul Aziz, Law Kiew Siong, Lau Song Lung, Preetvinder Singh a/l Dheer Singh, Tan Ying Sern, Eric Chung |
|---|---|
| Format: | Proceeding poster eJournal |
| Bahasa: | eng |
| Terbitan: |
, 2020
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/4008029 |
Daftar Isi:
- Introduction The radiographic appearance of Covid-19 infection is fairly unique – characterized by bilateral symmetrical ground-glass consolidation without pleural effusion – and has been reported on both chest CT and chest radiographs (CXRs). Based on the hypothesis that this uniqueness can be used to reliably diagnose Covid-19 on CXRs, we performed a study to assess and optimize the ability of COVID-Net, an open-source deep learning model published by the University of Waterloo, to detect Covid-19 on local CXRs. Objective To optimize and validate COVID-Net for the prediction of Covid-19 in CXRs of Malaysian patients Methods [Refer to Poster]. Results [Refer to Poster]. Conclusion The published (unoptimized) COVID-Net model has mediocre performance which improved tremendously after optimization. This suggests the model is robust but requires optimization with local CXRs before consideration for clinical use. Class activation maps do not reflect disease distribution. Further work is required for model explainability.