AI Solution: Inbody/Outbody detection

In today’s surgical documentation, more and more videos are being recorded thanks to modern endoscopic camera systems and improved documentation tools like our easyCapture solution. To share these recordings with patients or for scientific purposes, they must meet different criteria.

The most important consideration is the privacy of the patient and the operating room staff. While removing patient demographic data from the recorded videos is relatively easy (for example, using anonymization tools), it is also important to remove any non-surgical image sequences from the video itself.

Especially in endoscopy videos, there are often sequences recorded “out-body” — for example, when the endoscopy camera head is removed from inside the body. These segments may inadvertently capture the patient’s face or identifiable members of the operating room staff.

Removing such privacy-sensitive footage can be challenging, as it typically requires manual review of the entire video to identify and remove the out-body segments.

In cooperation with BOSTER, a project of University Hospital Bonn, a machine learning (AI) algorithm has been developed to analyze endoscopic videos and distinguish between in-body and out-body sequences on a frame-by-frame basis. This algorithm enables the automatic detection and removal or blurring of out-body sequences, making it easier to share edited videos with patients or for scientific purposes.

This inbody / outbody AI algorithm is available as an additional module within HCI ASMIS. The HCI ASMIS server is installed inside the hopitals  data centre, fulfilling the neccessary requirements for processing and storing patient related data. Thus, the inbody / outbody detection AI-algorithm is running inhouse, comletely deployed under hospital governance. Our AI algorithm operates without any internet or cloud connection, and there is no external party or company involved. The data of the patient remain the property of the patient. Commercial or intellectual property rights for recordings and images are not granted to anyone.