How Kepler Vision helps elderly facility patients get the right care

Preventing falls and other potentially harmful incidents among elderly and ailing patients is the responsibility of care facilities. Pull chords and body devices can cause false alarms, disturbing patients during their night’s rest and distracting caregivers from patients in need while adding stress to their shifts.

Kepler Vision Technologies develops artificial intelligence solutions that aid in healthcare using vision-based human activity recognition software. Its software known as the Kepler Night Nurse analyzes nighttime video streams of patients in elderly facilities to monitor if they need assistance and, if so, to alert the staff. Key to the high-end software is that it is intelligent enough to recognize real incidents versus false alarms. Headquartered in Amsterdam, the company has so far received 3.9 million euros in startup investment and currently employs 15 experts in machine learning, computer vision, and healthcare.

To accurately register patient falls within 60 seconds of occurrence – with a swift response sometimes proving to be a matter of life and death in such falls – Kepler Vision needed an IT solution with 100% uptime. The solution needed to be not only simple to implement, but also efficient in transmitting information to the appropriate caregivers and easy for them to use. At the same time, it needed to ensure patient data privacy and security.

To scale up its business in the Netherlands and internationally, Kepler Vision wanted a solution that could be quickly and securely scaled to process up to thousands of vision sensors per facility and at a growing number of facilities.

We created an IT solution for the Kepler Night Nurse that unifies the multiple and thus complicating forms of care facility automation that are often used within a single institution. Our solution deploys edge technology, meaning it runs on a device for local analysis and the data does not get transferred over the internet, but stays within a facility’s protected network. That, in turn, makes it simple to implement anywhere, even if a facility lacks the necessary bandwidth and technological basis. We could keep the Kepler Night Nurse up and running with relative ease and speed because the care facilities already used the public cloud and its ecosystem was equipped to integrate AI. Only sensors and software review the video stream images, which saves caregivers from having to do this work while also respecting patient privacy.

Whereas in the past, nearly 90% of alarm reports turned out to be false in a number of elderly facilities, locations using Kepler Vision with our solution expect the rate of alarm to fall from 3,000 to between just 60 and 80 per night. What’s more, the solution can scale data processing quickly and securely to potentially hundreds or thousands of sensors per institution. This solution has also helped increase the knowledge and experience of combining machine learning and artificial intelligence for mission-critical operations to software solutions developed for other care settings. It can also apply these learnings to other industries. For example, the vision-based human activity recognition software vision could monitor worker falls in heavy industry locations, where large machines and noise make it harder to report or detect injuries, or ensure that staff are wearing personal protective equipment.