How does the video de-identification work?

At Theator, ensuring the confidentiality and privacy of surgical video recordings is paramount. Our system uses advanced automation to remove any potentially identifiable information from surgical videos before they are stored or analyzed. Here’s how our automated de-identification processes work:

Automatic Removal of Identifiable Information:

1. Extracavitary Frame Removal:

   • Our technology automatically identifies and removes extracavitary (out-of-body) frames from surgical videos. These frames often contain identifiable features such as patient faces or skin marks.
   • For an in-depth look at how our technology detects out-of-body frames with high accuracy, refer to the research by Zohar et al., 2020, which shows a model accuracy of 99.3%. The accuracy have elevated even more since then. This research is detailed in the publication available here.

2. Metadata and Watermark Removal:

   • Metadata Removal: All video file metadata, such as dates and patient IDs, is automatically stripped from the recordings to prevent any accidental disclosure of identifiable information.
   • Watermark Removal: Names, IDs, or dates that are watermarked on the video are also automatically removed, including any on-screen overlays that could identify the patient or healthcare providers.

Real-Time, On-Premise De-identification:

• Immediate Processing: The de-identification process occurs in real-time, ensuring that no identifiable information ever leaves the operating room environment.
• On-Premise Security: All de-identification is performed on-premise (using Theator Unit) before the video stream is uploaded to the cloud. This approach not only enhances security but also ensures that the integrity and confidentiality of the data are maintained at all times.

By employing these advanced automated de-identification techniques, Theator ensures that all surgical videos are devoid of personal identifiers, thereby safeguarding patient privacy while facilitating the educational and analytical use of surgical videos without compromising confidentiality.

Further Reading:

For a deeper understanding of the standards and practices we follow, you may refer to the article "Defining the Standard for Surgical Video Deidentification" by Matthew K. Tollefson and Christopher J. Ross. Joint work between Theator and Mayo Clinic which discusses the importance and methods of surgical video de-identification in detail. This publication can be found in JAMA Surgery, 2023.

​