Technical Performance Assessment of Quantitative Imaging in Radiological Devices
This guidance is applicable to all devices that generate quantitative imaging values across radiological imaging modalities, intended uses, levels of automation, and complexity of algorithms. It provides FDA's recommendations on information, technical performance assessment, and user information that should be included in premarket submissions for radiological devices with quantitative imaging functions.
What You Need to Know? 👇
What are the key performance specifications required for quantitative imaging functions in FDA submissions?
FDA requires bias, precision, limits of quantitation, linearity, and uncertainty assessments. Performance specifications must cover the entire operating range and correspond to device claims. Supporting data should demonstrate the function meets predefined specifications under intended use conditions.
How does the level of automation affect validation requirements for quantitative imaging devices?
Fully automated functions require more robust analytical validation and uncertainty information than manual or semi-automated devices. Automated devices need testing on diverse clinical data representing expected use cases, while manual functions may rely more on software verification activities.
What technical documentation must be included when describing quantitative imaging algorithms to FDA?
Include measurand description with units, detailed algorithm inputs/outputs, level of automation, underlying physics for physics-based algorithms, training paradigm summary, input data restrictions, image acceptance activities, and user interaction requirements. Software documentation should follow FDA’s software guidance.
What are the essential labeling requirements for quantitative imaging functions?
Labeling must include measurand description and units, algorithm input restrictions, performance specifications with uncertainty information, quality assurance instructions, user qualification requirements, and reference database composition if applicable. Uncertainty information should be displayed on-screen when possible.
How should manufacturers address measurement uncertainty in quantitative imaging submissions?
Uncertainty must be assessed under conditions reflecting intended use and reported in measurand units with measurement conditions. It should cover the entire operating range since uncertainty may vary throughout. When specific metrics aren’t available, identify primary variability sources affecting outputs.
What types of reference standards are acceptable for validating quantitative imaging performance?
Phantoms with known ground truth values are preferred for objective comparison. Clinical images may be used but make accuracy characterization difficult. Reference materials, consensus values from collaborative work, or established values based on scientific principles are acceptable depending on the measurand.
What You Need to Do 👇
Recommended Actions
- Define and document the quantitative imaging function, measurand, and use conditions
- Determine appropriate reference standards and performance metrics
- Characterize performance under defined conditions
- Develop comprehensive performance specifications including uncertainty information
- Conduct appropriate verification and validation testing based on level of automation
- Prepare detailed labeling including:
- Technical descriptions
- Performance specifications
- User instructions
- Limitations and warnings
- Document sources of measurement error and their impact
- Ensure performance claims are supported by data
- Include appropriate quality control and acceptance protocols
- Consider level of automation when determining extent of validation required
Key Considerations
Non-clinical testing
- Technical performance assessment must demonstrate that specifications correspond to claims and uncertainty
- Performance testing should consider factors impacting performance (patient characteristics, image acquisition, image processing)
- Use of phantoms recommended as reference standards when possible
- Statistical analysis plan and acceptance criteria must be pre-defined
- More robust validation required for fully automated vs manual/semi-automated functions
Human Factors
- Description of qualifications and training needed for intended users
- Level of user interaction needed must be specified
- Instructions for image acceptance and quality assurance activities
Software
- Detailed description of algorithm, including inputs/outputs
- Software platform version and characteristics
- Level of automation (manual, automatic, semi-automatic)
- Algorithm training paradigm if applicable
- Software verification documentation
Labeling
- Description of measurand and units
- Algorithm inputs and restrictions
- Performance specifications including uncertainty information
- Instructions for quality assurance activities
- Reference database information if applicable
- Limitations and warnings
Other considerations
- Sources of measurement error must be identified and characterized
- Uncertainty assessment across operating range
- Performance specifications may vary throughout operating range
Relevant Guidances 🔗
- Technical Performance Assessment and Premarket Requirements for Digital Diagnostic Radiology Display Devices
- Content of Premarket Submissions for Device Software Functions
- Applying Human Factors Engineering and Usability Engineering to Medical Devices
- Harmonization of Performance Standards for Diagnostic X-Ray Imaging Systems and Components with IEC Standards
- Premarket Notification Requirements for Magnetic Resonance Diagnostic Devices
Related references and norms 📂
- JCGM 200:2012: International Vocabulary of Metrology – Basic and General Concepts and Associated Terms (VIM 3rd edition)