Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Detecting Performance Drift in AI Models for Medical Image Analysis Using CUSUM Chart.

Smriti Prathapan1, Ravi Samala2, Nathan Hadjiyski1

  • 1Office of Science and Engineering Laboratories, Center for Devices and Radiological Health, U.S. Food and Drug Administration, 10903 New Hampshire Ave, Silver Spring, 20993, MD, USA.

Journal of Imaging Informatics in Medicine
|July 14, 2026
PubMed
Summary

Related Concept Videos

Interpreting Run Charts01:25

Interpreting Run Charts

Run charts, essentially line graphs plotted over time, serve as fundamental yet effective tools for process analysis. They chronicle data sequentially, facilitating the identification of trends, shifts, or cyclical movements. This graphical representation is instrumental in determining whether a process is stable or exhibits signs of potential instability indicative of special cause variation. In the healthcare domain, run charts depict infection rates over time, enabling hospitals to monitor...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Physics-informed data augmentation to simulate low dose CT scans: Application to lung nodule detection.

Medical physics·2026
Same author

Network-based analysis of images and lesions expressed by digital interpretation and transmission of bronchoscopy findings on social media (NAILED-IT).

Respiratory medicine·2026
Same author

Synthetic data in radiological imaging: current state and future outlook.

BJR artificial intelligence·2026
Same author

Temporal Trends and Mortality in Tracheobronchial Stenting in the United States: a national cohort study.

Annals of the American Thoracic Society·2026
Same author

A miniature magnetic switch unlocking multimodal, chip-free, and batteryless airway sensing.

Science advances·2026
Same author

A Revolution in Robotic Bronchoscopy: Or Just Another 3-Body Problem?

Chest·2026

Monitoring artificial intelligence (AI) model performance in clinical settings is crucial. Cumulative sum (CUSUM) charts effectively detect data drift, ensuring reliable AI-assisted medical diagnosis and prediction.

Area of Science:

  • Medical Artificial Intelligence
  • Statistical Process Control
  • Clinical Decision Support Systems

Background:

  • Artificial intelligence (AI) models enhance clinical decisions but face performance degradation due to data drift post-deployment.
  • Monitoring deployed AI models is essential for ensuring expected performance and detecting deviations.

Purpose of the Study:

  • To investigate the efficacy of cumulative sum (CUSUM) control charts in detecting abrupt data drift affecting AI model performance.
  • To demonstrate CUSUM's application in monitoring AI for computer-aided breast cancer detection.

Main Methods:

  • Utilized CUSUM control charts to analyze performance metric changes in an AI model.
  • Applied the method to the Emory Breast Imaging Dataset (EMBED) for breast cancer detection.
Keywords:
Clinical AI workflowCumulative sum (CUSUM)Data drift detectionStatistical process control

Related Experiment Videos

  • Evaluated CUSUM's robustness to non-normal distributions and its ability to function without ground-truth labels.

    • Main Results:

    • CUSUM detected a 1.5 standard deviation performance drift within approximately 5 days, with a low false alarm rate (~293 days).
    • With 40 samples/day, CUSUM showed minimal false alarms over 60 days, balancing detection delay and false alarms.
    • CUSUM demonstrated robustness to non-normal data and effectiveness even without ground-truth labels.

    Conclusions:

    • CUSUM is a viable and adaptable statistical process control method for monitoring AI model performance in medical diagnosis.
    • The method ensures the reliability and expected performance of AI tools in clinical practice.
    • CUSUM facilitates early detection of performance deviations, crucial for patient safety and effective AI deployment.