This article is a Preprint
Preprints are preliminary research reports that have not been certified by peer review. They should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Preprints posted online allow authors to receive rapid feedback and the entire scientific community can appraise the work for themselves and respond appropriately. Those comments are posted alongside the preprints for anyone to read them and serve as a post publication assessment.
Accuracy of deep learning based computed tomography diagnostic system of COVID-19: a consecutive sampling external validation cohort study
Preprint
in English
| medRxiv
| ID: ppmedrxiv-20231621
Journal article
A scientific journal published article is available and is probably based on this preprint. It has been identified through a machine matching algorithm, human confirmation is still pending.
See journal article
A scientific journal published article is available and is probably based on this preprint. It has been identified through a machine matching algorithm, human confirmation is still pending.
See journal article
ABSTRACT
ObjectivesAli-M3, an artificial intelligence, analyses chest computed tomography (CT) and detects the likelihood of coronavirus disease (COVID-19) in the range of 0 to 1. It demonstrates excellent performance for the detection of COVID-19 patients with a sensitivity and specificity of 98.5 and 99.2%, respectively. However, Ali-M3 has not been externally validated. Our purpose is to evaluate the external validity of Ali-M3 using Japanese sequential sampling data. MethodsIn this retrospective cohort study, COVID-19 infection probabilities were calculated using Ali-M3 in 617 symptomatic patients who underwent reverse transcription-polymerase chain reaction (RT-PCR) tests and chest CT for COVID-19 diagnosis at 11 Japanese tertiary care facilities, between January 1 and April 15, 2020. ResultsOf 617 patients, 289 patients (46.8%) were RT-PCR-positive. The area under the curve (AUC) of Ali-M3 for predicting a COVID-19 diagnosis was 0.797 (95% confidence intervals [CI] 0.762-0.833) and goodness-of-fit was P = 0.156. With a cut-off of probability of COVID-19 by Ali-M3 diagnosis set at 0.5, the sensitivity and specificity were 80.6% and 68.3%, respectively, while a cut-off of 0.2 yielded a sensitivity and specificity of 89.2% and 43.2%, respectively. Among 223 patients who required oxygen support, the AUC was 0.825 and sensitivity at a cut-off of 0.5 and 0.2 were 88.7% and 97.9%, respectively. Although the sensitivity was lower when the days from symptom onset were few, sensitivity increased for both cut-off values after 5 days. ConclusionsAli-M3 was evaluated by external validation and shown to be useful to exclude a diagnosis of COVID-19. Key PointsO_LIThe area under the curve (AUC) of Ali-M3, which is an AI system for diagnosis of COVID-19 based on chest CT images, was 0.797 and goodness-of-fit was P = 0.156. C_LIO_LIWith a cut-off of probability of COVID-19 by Ali-M3 diagnosis set at 0.5, the sensitivity and specificity were 80.6% and 68.3%, respectively, while a cut-off of 0.2 yielded 89.2% and 43.2%. C_LIO_LIAlthough low sensitivity was observed in less number of days from symptoms onset, after 5 days high increasing sensitivity was observed. In patients requiring oxygen support, the AUC was higher that is 0.825. C_LI
cc_no
Full text:
Available
Collection:
Preprints
Database:
medRxiv
Type of study:
Cohort_studies
/
Diagnostic study
/
Experimental_studies
/
Observational study
/
Prognostic study
Language:
English
Year:
2020
Document type:
Preprint