Related Experiment Video
Updated: Jul 15, 2025
08:58
Artificial Intelligence Approaches to Assessing Primary Cilia
Published on: May 1, 2021
3.5K
Artificial intelligence-based image analysis in clinical testing: lessons from cervical cancer screening
Didem Egemen1, Rebecca B Perkins2, Li C Cheung1
1Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, MD, USA.
Journal of the National Cancer Institute
|September 27, 2023
Summary
Developing effective artificial intelligence (AI) screening tools requires rigorous validation and clear clinical goals. This approach ensures AI diagnostic tests translate into real-world clinical efficacy, especially in resource-limited settings.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Healthcare
- Clinical Diagnostics
Background:
- Artificial intelligence (AI) image recognition algorithms for screening and diagnostics are rapidly emerging.
- Initial claims of high accuracy for AI tests often lack confirmation, highlighting a gap between algorithm development and clinical utility.
Purpose of the Study:
- To present a conceptual, step-by-step framework for developing AI algorithms that achieve clinical efficacy.
- To guide the rigorous evaluation and implementation of AI-based medical diagnostic tests.
Main Methods:
- Clearly defining the AI algorithm's purpose (screening, diagnostic, prognostic) and target identification.
- Designing algorithms to minimize critical errors, including introducing an 'indeterminate' class for equivocal cases.
- Evaluating AI performance using clinical epidemiologic criteria, distinguishing internal and external validation.
- Linking AI results to absolute clinical risk estimation and developing risk-based guidelines.
Main Results:
- A structured approach is proposed to bridge the gap between AI algorithm creation and clinical efficacy.
- The importance of defining clear objectives, managing equivocal data, and using robust validation methods is emphasized.
- Translating AI test results into clinical practice requires focusing on absolute risk and context-specific guidelines.
Conclusions:
- A systematic methodology is crucial for ensuring the reliability and clinical utility of AI diagnostic tools.
- This framework is particularly relevant for addressing health disparities in lower-resource settings.
- The principles discussed are applicable to various AI-based medical image analysis applications.
