Is the generalizability of a developed artificial intelligence algorithm for COVID-19 on chest CT sufficient for clinical use? Results from the International Consortium for COVID-19 Imaging AI (ICOVAI) - Summary - MDSpire

Is the generalizability of a developed artificial intelligence algorithm for COVID-19 on chest CT sufficient for clinical use? Results from the International Consortium for COVID-19 Imaging AI (ICOVAI)

  • By

  • Laurens Topff

  • Kevin B. W. Groot Lipman

  • Frederic Guffens

  • Rianne Wittenberg

  • Annemarieke Bartels-Rutten

  • Gerben van Veenendaal

  • Mirco Hess

  • Kay Lamerigts

  • Joris Wakkie

  • Erik Ranschaert

  • Stefano Trebeschi

  • Jacob J. Visser

  • Regina G. H. Beets-Tan

  • January 18, 2023

  • 0 min

Share

Objective:

To develop and independently validate a specific AI model for COVID-19 segmentation and CO-RADS classification on chest CT using multicenter data.

Key Findings:
  • The AI model demonstrated reasonable accuracy in segmenting lung tissue and classifying CO-RADS scores, suggesting potential for clinical integration.
  • External validation indicated the model's generalizability across different patient populations and imaging conditions, highlighting its applicability in diverse clinical settings.
  • Automated CO-RADS classification could enhance clinical workflow and interobserver agreement, particularly benefiting less experienced radiologists.
Interpretation:

The AI model shows promise for clinical application in COVID-19 imaging, but further validation is necessary to ensure reliability in diverse clinical settings, including various demographics and imaging conditions.

Limitations:
  • The study's dataset may not represent all demographic and clinical variations encountered in real-world scenarios, potentially limiting the model's applicability.
  • Potential biases in data collection and reader experience, such as variability in radiologist expertise, may affect the model's performance.
Conclusion:

The developed AI model for COVID-19 imaging has potential for clinical use, pending further validation to confirm its generalizability and accuracy.

Original Source(s)

Is the generalizability of a developed artificial intelligence algorithm for COVID-19 on chest CT sufficient for clinical use? Results from the International Consortium for COVID-19 Imaging AI (ICOVAI)

  • by Laurens Topff, Kevin B. W. Groot Lipman, Frederic Guffens, Rianne Wittenberg, Annemarieke Bartels-Rutten, Gerben van Veenendaal, Mirco Hess, Kay Lamerigts, Joris Wakkie, Erik Ranschaert, Stefano Trebeschi, Jacob J. Visser, Regina G. H. Beets-Tan

  • January 18, 2023

Related Content

Deep-Motion-Net: GNN-based volumetric liver shape reconstruction from single-view 2D projections

Epidemiology of respiratory syncytial virus in a German hospital before, during and after the COVID-19 pandemic – a real world monocentric analysis over ten years

Combining radiomics and machine learning for enhanced localization of premature ventricular contractions