Dataset Information

Comparison of the Diagnostic Performance of Deep Learning Algorithms for Reducing the Time Required for COVID-19 RT-PCR Testing.

ABSTRACT: (1) Background: Rapid and accurate negative discrimination enables efficient management of scarce isolated bed resources and adequate patient accommodation in the majority of areas experiencing an explosion of confirmed cases due to Omicron mutations. Until now, methods for artificial intelligence or deep learning to replace time-consuming RT-PCR have relied on CXR, chest CT, blood test results, or clinical information. (2) Methods: We proposed and compared five different types of deep learning algorithms (RNN, LSTM, Bi-LSTM, GRU, and transformer) for reducing the time required for RT-PCR diagnosis by learning the change in fluorescence value derived over time during the RT-PCR process. (3) Results: Among the five deep learning algorithms capable of training time series data, Bi-LSTM and GRU were shown to be able to decrease the time required for RT-PCR diagnosis by half or by 25% without significantly impairing the diagnostic performance of the COVID-19 RT-PCR test. (4) Conclusions: The diagnostic performance of the model developed in this study when 40 cycles of RT-PCR are used for diagnosis shows the possibility of nearly halving the time required for RT-PCR diagnosis.

SUBMITTER: Lee Y

PROVIDER: S-EPMC9966023 | biostudies-literature | 2023 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Comparison of the Diagnostic Performance of Deep Learning Algorithms for Reducing the Time Required for COVID-19 RT-PCR Testing.

Lee Yoonje Y Kim Yu-Seop YS Lee Da In DI Jeong Seri S Kang Gu Hyun GH Jang Yong Soo YS Kim Wonhee W Choi Hyun Young HY Kim Jae Guk JG

Viruses 20230122 2

(1) Background: Rapid and accurate negative discrimination enables efficient management of scarce isolated bed resources and adequate patient accommodation in the majority of areas experiencing an explosion of confirmed cases due to Omicron mutations. Until now, methods for artificial intelligence or deep learning to replace time-consuming RT-PCR have relied on CXR, chest CT, blood test results, or clinical information. (2) Methods: We proposed and compared five different types of deep learning ...[more]

PMID: 36851519

Similar Datasets

Project description:BackgroundThe comparative performance of different clinical sampling methods for diagnosis of SARS-CoV-2 infection by RT-PCR among populations with suspected infection remains unclear. This meta-analysis aims to systematically compare the diagnostic performance of different clinical specimen collection methods.MethodsIn this systematic review and meta-analysis, we systematically searched PubMed, Embase, MEDLINE, Web of Science, medRxiv, bioRxiv, SSRN, and Research Square from Jan 1, 2000, to Nov 16, 2020. We included original clinical studies that examined the performance of nasopharyngeal swabs and any additional respiratory specimens for the diagnosis of SARS-CoV-2 infection among individuals presenting in ambulatory care. Studies without data on paired samples, or those that only examined different samples from confirmed SARS-CoV-2 cases were not useful for examining diagnostic performance of a test and were excluded. Diagnostic performance, including sensitivity, specificity, positive predictive value, and negative predictive value, was examined using random effects models and double arcsine transformation.FindingsOf the 5577 studies identified in our search, 23 studies including 7973 participants with 16 762 respiratory samples were included. Respiratory specimens examined in these studies included 7973 nasopharyngeal swabs, 1622 nasal swabs, 6110 saliva samples, 338 throat swabs, and 719 pooled nasal and throat swabs. Using nasopharyngeal swabs as the gold standard, pooled nasal and throat swabs gave the highest sensitivity of 97% (95% CI 93-100), whereas lower sensitivities were achieved by saliva (85%, 75-93) and nasal swabs (86%, 77-93) and a much lower sensitivity by throat swabs (68%, 35-94). A comparably high positive predictive value was obtained by pooled nasal and throat (97%, 90-100) and nasal swabs (96%, 87-100) and a slightly lower positive predictive value by saliva (93%, 88-97). Throat swabs have the lowest positive predictive value of 75% (95% CI 45-96). Comparably high specificities (range 97-99%) and negative predictive value (range 95-99%) were observed among different clinical specimens. Comparison between health-care-worker collection and self-collection for pooled nasal and throat swabs and nasal swabs showed comparable diagnostic performance. No significant heterogeneity was observed in the analysis of pooled nasal and throat swabs and throat swabs, whereas moderate to substantial heterogeneity (I2 ≥30%) was observed in studies on saliva and nasal swabs.InterpretationOur review suggests that, compared with the gold standard of nasopharyngeal swabs, pooled nasal and throat swabs offered the best diagnostic performance of the alternative sampling approaches for diagnosis of SARS-CoV-2 infection in ambulatory care. Saliva and nasal swabs gave comparable and very good diagnostic performance and are clinically acceptable alternative specimen collection methods. Throat swabs gave a much lower sensitivity and positive predictive value and should not be recommended. Self-collection for pooled nasal and throat swabs and nasal swabs was not associated with any significant impairment of diagnostic accuracy. Our results also provide a useful reference framework for the proper interpretation of SARS-CoV-2 testing results using different clinical specimens.FundingHong Kong Research Grants Council.

Dataset Information

Comparison of the Diagnostic Performance of Deep Learning Algorithms for Reducing the Time Required for COVID-19 RT-PCR Testing.

Publications

Comparison of the Diagnostic Performance of Deep Learning Algorithms for Reducing the Time Required for COVID-19 RT-PCR Testing.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets