Do I Need to Validate My Imaging Study?
Validating every step of your imaging studies can help you maintain data integrity, reproducibility and achieve regulatory compliance with 21 CFR Part 11. But whether you need to validate your imaging study depends on the specific needs of your organization. Below, we outline common scenarios across organizations like academic medical centers, pharmaceutical companies and other research enterprises to help you determine whether validation is necessary.
1. I’m an academic medical center or hospital doing internal research to improve delivery of care.
Answer: No
If your imaging study is for internal quality improvement or exploratory research and not intended for regulatory submission, formal FDA approval isn't required. That only changes with regard to imaging devices and software, and if they’re used in patient care.
For example, if a radiology department uses a device that displays imaging data to optimize MRI scheduling protocols, that wouldn’t be subject to FDA oversight. However, devices or software that impact clinical decision-making do need to be FDA-approved.
2. I'm an academic medical center conducting NIH-funded clinical research.
Answer: Maybe
NIH-funded studies involving human subjects must comply with other regulations, including 45 CFR Part 46 and sometimes 21 CFR Part 50. It’s only when these studies involve FDA-regulated products and electronic records and signatures that must be submitted to the FDA or are used in regulated activities that they need to comply with 21 CFR Part 11.
3. I'm a pharmaceutical company executing imaging studies in support of novel drug-based therapies.
Answer: Yes
Pharmaceutical companies conducting clinical trials must validate imaging studies used to support regulatory submissions. The FDA and European Medicines Agency (EMA) also require that novel imaging biomarkers used as endpoints or for patient stratification be validated.
4. I'm a contract research organization (CRO) managing imaging for a sponsor’s clinical trial.
Answer: Yes
CROs must ensure that imaging data collected and analyzed on behalf of sponsors meets regulatory and scientific standards. This includes validation of imaging acquisition, quality check, anonymization, transfer, archive, quantitative image analysis and independent blinded image review.
For example, a CRO managing imaging endpoints for a Phase III oncology trial sponsored by a pharmaceutical company needs to validate its imaging workflows to support FDA submission.
5. I'm a medical device company developing AI-based imaging tools to integrate into our instruments.
Answer: Yes
AI tools that analyze imaging data for diagnostic or therapeutic purposes are considered medical devices and must undergo validation. In fact, that validation may be more rigorous for devices that include AI capabilities, requiring developers to identify and mitigate risks, such as bias, and to ensure devices are validated in a realistic clinical setting so their output is clinically meaningful.
Say, if a startup is developing an AI algorithm to detect lung nodules on CT scans, validation is required to demonstrate the tool’s safety and effectiveness before FDA clearance.
6. I'm a hospital IT team implementing a new imaging platform for clinical use.
Answer: Yes (mostly)
When hospitals use imaging platforms in their clinical workflows, especially those that interface with PACS or EHR systems, those platforms must be validated. They’re only exempt if they’re solely used to transfer data, store data, convert formats, or display medical device data and results.
7. I'm a research institution using public imaging datasets for exploratory analysis.
Answer: No (but be cautious)
If you’re using publicly available datasets for hypothesis generation or exploratory research, you don’t need formal validation. So, if you’re using ADNI dataset to explore machine learning models for dementia classification, that dataset is pre-validated and doesn’t require additional validation. However, once your research progresses to the place in which you want to conduct clinical trials, your workflows, devices and newly collected data will need validation.
8. I’m a student testing a hypothesis with public datasets.
Answer: No (but be cautious)
You’re safe here — even though the data you’re using came from human beings, they don’t qualify as “human subject research” because that data has been de-identified, and there’s no interaction with any individual involved. Just like in the above example, you’ll need to be careful if you merge multiple datasets in a way that enables identification of the individuals whose data is being analyzed.
9. I’m a pharmaceutical company testing a hypothesis prior to a clinical trial.
Answer: No, unless you want to submit your findings
For exploratory research that won’t be part of an official clinical trial submitted to the FDA, you don’t need a validated workflow. But you may at the very least need an audit trail if during the course of that research you want to recreate your findings or later possibly submit them to the FDA for approval.
For example, if a biotech firm is trialing ideal candidates for a new radiopharmaceutical for prostate cancer, imaging data gathered and compared to assess tumor response doesn’t need to be validated. However, if you end up wanting to use that data for a submission to the FDA, you will need an audit trail and to follow 21 CFR Part 11 compliance.
10. I don’t know if my study qualifies as a clinical trial or not.
Answer: You need to find out
Unfortunately, you can’t claim ignorance on this one. It’s incumbent on the entity conducting the study to determine if it qualifies as a clinical trial subject to FDA approval. It’s worth reviewing what the NIH defines as a clinical trial and then adjusting your research capabilities as needed.
Making Validation Easier
Whether you're conducting internal research or preparing for regulatory submission, the need for validation depends on your study’s purpose and how you’re handling data. Regardless of your validation requirements, Flywheel can help you make the most of your imaging data by streamlining imaging data management, automating research workflows and preparing data for AI development. And Flywheel has several configurations to fit your organization’s needs.
Flywheel Core is ideal for research teams that need a powerful, flexible platform to manage imaging data across the enterprise and support collaboration between institutions and locations.
Audit Trail capabilities can also be added to Flywheel Core if you need enhanced tracking for AI development, imaging data management or full data provenance for collaboration.
If you need a fully validated environment for frequent FDA submission, Flywheel Validated Core helps you meet 21 CFR Part 11 compliance while retaining more control over your data so you can unlock secondary uses, experiment with hypotheses and optimize your clinical trials. It includes automated, validated audit trails; access controls; and the ability to lock data subject to audit or submission to a regulatory body.
No matter your use case, Flywheel can help you manage imaging data with confidence. Book a demo today to learn which option will help you reach your research goals and achieve compliance when needed.
