A Deep Learning System to Screen Novel Coronavirus Disease 2019 Pneumonia.
Source check: Source checked. Not independent scientific confirmation.
Finding: The study distinguished COVID-19 from influenza-A viral pneumonia and healthy cases; overall accuracy was 86.7% on a held-out set of 90 CT cases (30 per category). The source reports 618 CT samples from three hospitals; 528 samples were used for training/validation and 90 were held out from patients not included in training. The test is an internal held-out sample, not an independent external cohort. Published abstract/full text do not establish clinical benefit or prospective utility.
Technology: A 3D deep-learning model segmented candidate infection regions; a location-attention classifier categorized CT regions and a Noisy-or Bayesian function combined confidence into a case-level prediction.
Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.
- AI contribution
- A 3D deep-learning model segmented candidate infection regions; a location-attention classifier categorized CT regions and a Noisy-or Bayesian function combined confidence into a case-level prediction.
- Test and result
- The study distinguished COVID-19 from influenza-A viral pneumonia and healthy cases; overall accuracy was 86.7% on a held-out set of 90 CT cases (30 per category).
- Limitation / still unknown
- The source reports 618 CT samples from three hospitals; 528 samples were used for training/validation and 90 were held out from patients not included in training. The test is an internal held-out sample, not an independent external cohort. Published abstract/full text do not establish clinical benefit or prospective utility.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/abs/2002.09334v1 · https://www.sciencedirect.com/science/article/pii/S2095809920301636 · https://pmc.ncbi.nlm.nih.gov/articles/PMC7320702/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: The first located result-bearing version is the authors’ arXiv v1 dated 2020-02-21; its abstract reports the same 86.7% CT-case accuracy later reported in the journal article. The original full-text study is a specific retrospective diagnostic-validation question, eligible under the project criteria. Active/archived catalog searches found no same-cohort record; the active Mount Sinai COVID-CT entry is a different model, input set and cohort. A later Applied Intelligence article was retracted for reusing this study’s data/text and is treated as the same underlying work, not a separate discovery..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.