De-identify clinical trial video before it leaves the study team
One-click alternative with BGBlur
Upload a clip and preview automatic face, plate, background, or prompt-based blur—no keyframes or nested timelines.
Video shows up throughout clinical research in ways a paper case-report form never captures: behavioral-observation coding for a pediatric or psychiatric trial, video-recorded patient-reported outcome diaries, gait or motor-function endpoints, procedural documentation for a device trial, and recruitment or training material built from real participant footage. Once that footage needs to travel — to a sponsor's central reviewers, into an IRB submission packet, into a publication's supplementary materials, or into a multi-site training set — the participant's identity becomes a liability the original informed-consent form may not have covered.
Most study consent language addresses use of video within the study team and, sometimes, de-identified use in publication — it rarely authorizes an identifiable participant appearing in sponsor training decks or public supplementary data. De-identifying the footage before it crosses that boundary keeps the study within what participants actually agreed to, and it's usually far faster than re-consenting a cohort for broader use.
Where video de-identification fits in the protocol
Treat video the same way your protocol already treats other identifiable data: define at protocol-design stage which video will be collected, who sees the identifiable version, and what de-identification step happens before video crosses into a broader-access dataset (sponsor server, publication supplement, multi-site archive). Building the blur step into the data-management plan avoids the common failure mode where a coordinator improvises redaction under deadline pressure before a submission.
IRBs reviewing a protocol with video endpoints will typically ask exactly this: what happens to identifiable footage before it leaves the site. Having a documented, repeatable de-identification step — not a promise to 'edit it later' — is usually the difference between a straightforward approval and a stipulation requiring revision.
Behavioral coding and endpoint video without losing the signal
Behavioral-observation and motor-endpoint studies depend on exactly the movement and expression detail that face blur is designed to preserve outside the masked region — coders working from de-identified footage can still score gaze direction, affect, gross motor patterns, and interaction behavior; blurring removes recognizability, not the behavioral signal itself, for most standard coding schemes. For protocols where fine-grained facial-expression coding (e.g., FACS-based affect coding) is the actual endpoint, discuss with your PI and biostatistician whether de-identification should happen after primary coding is complete, since blurring would remove the data being measured.
For studies where the face itself isn't the endpoint — gait, medication-adherence video diaries, wound documentation, procedural technique — blur-before-distribution is straightforward and doesn't touch the primary outcome data at all.
- Motor/gait/procedural endpoints: blur before any distribution beyond the site.
- Affect or facial-expression coding endpoints: coordinate timing with your PI/biostatistician.
- Video diaries and PRO recordings: blur before sponsor or central-reader access unless the protocol specifies otherwise.
Sponsor sharing, publication, and multi-site archives
Sponsors and CROs often want video access for central review, adjudication committees, or quality audits — access that's easier to grant when the footage is de-identified, since it reduces the data-transfer and access-control burden on both sides. A standard batch pass before each transfer keeps this from becoming a bottleneck at every sponsor request.
For publication, most journals' supplementary-data policies require identifiable participant material to be either explicitly consented for publication or de-identified. Running the final supplementary clips through a blur pass before submission is generally simpler than chasing publication-specific consent from every participant whose footage appears, particularly in trials that closed enrollment years earlier.
From site recording to submission-ready footage
- Export the site recording. Save the behavioral, endpoint, or procedural clip as recorded — no re-editing needed first.
- Confirm the de-identification point. Check with your PI whether coding happens before or after blurring for facial-expression endpoints.
- Upload and blur. Automatic face detection masks participants through the full clip, including movement and re-entries into frame.
- Batch across the cohort. Apply the same pass across multiple participants' clips before a sponsor transfer or submission deadline.
- Export and archive. Deliver the de-identified derivative to the sponsor, IRB packet, or publication supplement; retain the identifiable original per your data-retention plan.
Note: The Common Rule (45 CFR 46) and most IRBs distinguish identifiable from de-identified research data, and video with a visible participant face is generally treated as identifiable. Confirm your specific de-identification standard with your IRB before relying on face blur alone to change a study's identifiability classification.
Related guides
Frequently asked questions
- Does blurring participant faces satisfy IRB de-identification requirements on its own?
- It addresses the visual identifier, which is usually the main concern for video endpoints, but de-identification under the Common Rule (45 CFR 46) or your IRB's standard may also require handling names, dates, MRNs, and other identifiers in metadata or accompanying documentation. Confirm the full de-identification standard your IRB expects and apply blur as one component of it — not a substitute for the rest.
- Will de-identifying the video affect our endpoint data?
- For motor, procedural, or gait endpoints, no — the movement signal is unaffected outside the masked facial region. For studies where facial-expression coding is the primary endpoint, coordinate timing with your PI so coding happens on the identifiable footage before de-identified copies are distributed further.
- Do we need separate consent to use blurred footage in a publication?
- Journal policies vary, but de-identified footage generally faces a lower consent bar than identifiable participant video. Check your target journal's specific supplementary-data and consent policy — this varies enough by publisher that it's worth confirming before submission.
- Can we process an entire cohort's footage before a sponsor deadline?
- Yes — batch processing applies the same blur pass across a folder of participant clips, which is the practical way to clear a cohort's footage ahead of a central-review or database-lock deadline.
- Is the original identifiable footage preserved after processing?
- Processing produces a new de-identified file; the source recording isn't altered in place. Keep the identifiable original under your protocol's data-retention and access-control plan — the de-identified copy is what travels to broader-access audiences.
BGBlur provides privacy tooling for creators and teams; consult counsel for broadcast, evidentiary, or regulated workflows.