Post

Clinical Performance Assessment of Computer-Assisted Detection Devices in Radiology

This guidance provides recommendations for clinical performance assessment of computer-assisted detection (CADe) devices applied to radiology images and data. It covers CADe devices that identify, mark, highlight, or direct attention to portions of radiological images/data that may reveal abnormalities during interpretation by clinicians. The guidance applies to devices classified under 21 CFR 892.2050 and 892.2070, including those for mammography, ultrasound, radiograph lung nodules, and dental caries.

  1. Develop detailed clinical study protocol with pre-specified endpoints and analysis plan
  2. Implement MRMC study design with appropriate controls and randomization
  3. Define and document reference standard methodology
  4. Establish reader training program
  5. Create audit trail system if planning test data reuse
  6. Document potential sources of bias and mitigation strategies
  7. Prepare comprehensive results reporting including subgroup analyses
  8. Consider requesting FDA protocol review through Pre-Submission
  9. Ensure data poolability across collection sites
  10. Maintain strict data integrity controls throughout assessment process

Key Considerations

Clinical testing

  • Must demonstrate clinical safety and effectiveness through well-controlled studies
  • Multiple Reader Multiple Case (MRMC) study design recommended
  • Study endpoints should include ROC-based metrics and sensitivity/specificity
  • Need adequate sample size and statistical power
  • Must include appropriate range of normal and abnormal cases
  • Reading sessions should be separated by at least 4 weeks
  • Training of readers required on device use and rating scales

Human Factors

  • Must train readers on device use and study participation
  • Reading scenarios should match intended clinical use
  • Need to consider impact of reading order and context bias

Labelling

  • Must specify intended use population
  • Must describe reading methodology consistent with clinical practice
  • Need to specify training requirements

Safety

  • Must demonstrate safety through clinical performance assessment
  • Need to control for potential sources of bias
  • Must maintain data integrity if reusing test data

Other considerations

  • 21 CFR 892.2050: Medical image management and processing system
  • 21 CFR 892.2070: Medical image analyzer
  • 21 CFR 814.15: Medical device approval based on foreign data

Original guidance

  • Clinical Performance Assessment of Computer-Assisted Detection Devices in Radiology
  • HTML / PDF
  • Issue date: 2022-09-28
  • Last changed date: 2022-09-27
  • Status: FINAL
  • Official FDA topics: Radiation-Emitting Products, Medical Devices, Digital Health, 510(k), Premarket, Labeling, Radiology
  • ReguVirta summary file ID: 58b8913b96761333f15c89cf68094d03
This post is licensed under CC BY 4.0 by the author.