The company claims its system cuts 99 per cent of false alarms in care homes

Dutch medtech startup Kepler Vision Technologies has launched an edge device that uses computer vision to improve elderly care.

The Kepler Night Nurse Edge Box analyzes live video feeds in care homes and hospitals to detect when patients are in distress or in need assistance, and alert staff immediately.

After the initial training period, the system can handle the video feeds so they never again have to be seen by a human being - and the video data doesn’t have to travel outside the facility.

The all-seeing eye

The Kepler Night Nurse AI system was developed by Dr Harro Stokman at the University of Amsterdam (UoA), using deep learning to make sense of the patterns of events in care homes.

Since the project was launched in 2018, the pandemic has intensified pressure on care homes, with high infection rates and staff shortages, while privacy laws have restricted the use video surveillance.

In response, UoA spin off Kepler Vision Technologies (KVT) used Nvidia’s small form factor Jetson Xavier NX module to build an edge computing device. The Edge Box can process data locally, saving on multiple connections to the cloud, while simultaneously reducing the rate of false alarms produced by motion detectors and sensors buried in bed mats, necklaces, and bracelets.

After a three-month calibration period the Edge Box improves patient privacy by taking over surveillance from staff. It informs them by text if patients need help.

The software includes a proprietary body language recognition system that can tell acceptable physical movement from distressed body language or fall. Behavioral observations – such as struggles or bathroom dawdling – can also be added to a patient’s medical file by the Night Nurse.

KVT has applied for 17 patents to date, and received €3.9 million from UvA Ventures and the Netherlands Ministry of Economic Affairs.