Download This Paper
Preprints with The Lancet is a collaboration between The Lancet Group of journals and SSRN to facilitate the open sharing of preprints for early engagement, community comment, and collaboration. Preprints available here are not Lancet publications or necessarily under review with a Lancet journal. These preprints are early-stage research papers that have not been peer-reviewed. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. The findings should not be used for clinical or public health decision-making or presented without highlighting these facts. For more information, please see the FAQs.
Free AI Software for Automatic CT Quantification of Coronavirus Disease 2019: An International Collaborative Development, Validation, and Distribution
42 Pages Posted: 11 Aug 2020
More...Abstract
Background: We aimed to develop and distribute free artificial intelligence (AI) software for the automatic quantification of coronavirus disease 2019 (COVID-19) pneumonia on computed tomography (CT) images.
Methods: We included 150 chest CT scans (17 CT scanners; five vendors) of 105 COVID-19 patients from 13 Korean and Chinese institutions. Two experienced radiologists semi-automatically drew lung opacities in every CT image, preparing 28,580 positive and negative CT slices to develop the 2D U-Net for segmenting pneumonia. The 2D U-Net was distributed as downloadable free software for local use on computers without data privacy concerns. External validation was performed using a Japanese single-institutional dataset and a public Italian dataset. Primary measures for the performance of the network were correlation coefficients for extent (%) and weight (g) of pneumonia. We surveyed user experiences.
Findings: In the internal validation dataset, the intraclass correlation coefficients between the 2D U-Net and reference values for the extent and weight were 0·987 and 0·992, respectively. In the Japanese dataset, the Pearson correlation coefficients between the visual CT severity score and 2D U-Net outcomes were 0·906 and 0·916, respectively. In the Italian dataset, the intraclass correlation coefficients between the 2D U-Net and reference values for extent and weight were 0·921 and 0·970, respectively. The median satisfaction score was 9/10 and most respondents replied to use the software willingly.
Interpretation: AI software for the automatic quantification of COVID-19 pneumonia on CT images was successfully developed and distributed freely worldwide.
Funding Statement: There was no funding source for this study.
Declaration of Interests: Sang Joon Park is the CEO of Medical IP. Gin Mo Goo has grants from Infinitt Healthcare, grants from Dongkook Lifescience, outside the submitted work. All the other authors have no potential conflicts of interest to disclose.
Ethics Approval Statement: The institutional review board of the participating hospitals approved this retrospective study, and the requirement for patient consent was waived.
Keywords: COVID-19; Pneumonia; Computed tomography; Deep learning
Suggested Citation: Suggested Citation