引導式測驗 · 中等 等級

醫療 AI 的臨床驗證測驗

測試分析、臨床和實用驗證;人口契合度;資料外洩;監控;以及回顧性表現的限制。

相關引導路徑醫療 AI 的臨床驗證
問題 1 的 8

What does clinical validation assess?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. What does clinical validation assess?
    • Whether code compiles
    • Whether a model has many parameters
    • Whether the vendor has a quality certificate
    • Whether outputs are relevant to the clinical state or decision in the intended setting
  2. How does analytical validation differ from clinical validation?
    • It checks input processing rather than clinical association or use
    • They are the same metric
    • Analytical validation proves patient benefit
    • Clinical validation checks only software installation
  3. Does retrospective test performance prove clinical utility?
    • Yes, if data were de-identified
    • No retrospective study can measure accuracy
    • No; utility requires evidence that use improves decisions or outcomes
    • Yes, if accuracy is high
  4. Why test on external sites or cohorts?
    • It changes the model’s intended purpose
    • Inputs, populations, and workflows may differ from development data
    • External testing guarantees every site performs equally
    • It removes the need for local evaluation
  5. Which split creates a data-leakage risk?
    • A model uses an independent test cohort
    • A report contains missing data
    • A threshold is prespecified
    • Related records from one patient appear in both training and test sets
  6. What should be evaluated beyond accuracy?
    • Only vendor response time
    • Only number of predictions
    • Calibration, subgroup performance, usability, and workflow effects
    • 僅型號尺寸
  7. When might new validation be needed?
    • When the model, inputs, population, or workflow changes
    • Only when a new logo is released
    • Never after initial testing
    • Only after changing the study title
  8. What does the IMDRF SaMD framework say about clinical evaluation?
    • It can be replaced by user satisfaction
    • It is iterative and includes association, analytical, and clinical validation evidence
    • It is a one-time software certification
    • It applies only to non-medical apps