A step-by-step plan for integrating the Kepler Night Nurse into the IT ecosystem of healthcare institutions

After you signed a deal with Kepler Vision Technologies to install the Kepler Night Nurse, there is a fixed step-by-step plan to successfully integrate our software into your IT ecosystem. The deal made could be a direct agreement between Kepler Vision Technologies and you or may be concluded with the intervention of one of our integration partners. These integration partners help with the installation of the required infrastructure and with the integration of the Kepler Night Nurse with your existing nurse call system. You will find a few partners with whom we regularly work alongside.

Below is a diagram of the steps that are taken for integration. These are the same for all packages of the Kepler Night Nurse.

integration steps for the Kepler Night Nurse

Figure 1: Scheme Method

Kick-off meeting

First, a kick-off meeting is held with all involved. The Kepler Vision Technologies project leader presents the action plan, and both parties propose a contact person so that both parties know who is responsible for the successful execution of the project. For Kepler, this will often be the project leader.


Step 1: Optical sensor installation

The Kepler Night Nurse works with live images from optical sensors, so these must be installed in the rooms that are being monitored. The brand of the optical sensor does not matter, as long as they meet the following conditions:


 Dome camera Fish Eye 
 4 MP or higher 6 MP of higher
 Nightvision Nightvision
 Wide Dynamic Range Wide Dynamic Range
 Horizontal FOV> 100º (2,8 mm lens) 360º
 Support RTSP, 2 tot 4 MBit/s Support RTSP, 4 tot 6 MBit/s
 Constant bit rate Constant bit rate
 Support uploading image by URL  Support uploading image by URL 
 Codec H264 or H265 Codec H264 or H265

The Kepler Night Nurse would also function using optical sensors with poorer specifications. However, this will be at the expense of the recognition accuracy of our software

The optical sensors must be able to oversee the monitored spaces as much as possible. This is important because otherwise, clients can “disappear,” for example behind a cupboard or bed lift. This could lead to false alarms, in this case, an “out of room” alarm, which should be prevented. Two types of optical sensors are suitable for this, namely:

  1. Dome type sensors with an angle of view of at least 100º horizontally / 58º vertically
  2. 360º optical sensors (fisheye).

The former is suitable when the optical sensor is installed in the corner of the room. The latter when it is installed more towards the center of the ceiling.

Next, Steps 2A and 2B are taken simultaneously:


Step 2A: Data collection

The Kepler Night Nurse software analyzes live images from optical sensors and is based on machine learning models. These models are calibrated and finetuned to the images at the locations of our customers. That is why it is necessary to collect and store so-called training and test data. This data is used to train the Kepler Night Nurse to location specific data. During this step, the images don’t need to be available ”live’’ for Kepler. Prerecorded data is sufficient for this step.

The data collection is divided into two parts.

1. Recordings with employees of Kepler (and possibly the customer).

The scenarios that the Kepler Night Nurse software must detect as events are played out. As a result, many events can be recorded in a short time such as falls – these are scarce with real clients. This way about three hours of video will be recorded. For this, Kepler gets at least 3 hours of access to selected bedrooms at the customer’s location from which images can be recorded. 100% of this data is then stored by Kepler Vision Technologies.

2. Recordings with clients.

We then collect at least 100 hours of video material from rooms and clients in a real situation. To achieve this, separate permission is required from the clients, family, or employees involved. From this data set, approximately 1% is used and stored. In the remaining 99%, nothing happens that is relevant to training.

Step 2B: Setting up a safe video connection between client and Kepler Night Nurse

A secure video connection is established between your nursing home and the cloud where we run the Kepler Night Nurse software. Figure 2 shows a schematic overview of the integration between customers and the Kepler Night Nurse. The result of this step is that images are sent via a secure connection (VPN) to Kepler Night Nurse. The Kepler Night Nurse analyzes these images. When an event is detected, it sends a message to the nurse call system, again via VPN. The nurse call system routes the message routed further within your IT ecosystem. This way your client data is always safe. If no nurse call system is present, a solution will be sought for this.


Integratie schema

Figuur 2: Overview implementation calibration step

Step 3: Calibration and assurance of the recognition accuracy of the Kepler Night Nurse

This step takes about four weeks and sometimes a little longer. The Kepler Night Nurse software is further calibrated on the live images from the customer’s location and the recognition accuracy is determined. This step can only be started once the data collection and integration steps have been completed.

In this step, the software may not have seen enough images of the specific situations to achieve the recognition accuracy desired by Kepler.

During the measurement of the recognition accuracy, the correct functioning of the Kepler Night Nurse is ensured with the help of video streams from the location of the customer. Both the real incidents and the generated alarms are measured. All situations fall into three categories:

  • Real situations where the Kepler Night Nurse rightly raised the alarm.
  • Real situations where the Kepler Night Nurse did not sound the alarm.
  • Situations where the Kepler Night Nurse sounded a false alarm.

These measurements are also done for the existing solution (if it is in place) so that a good comparison can be made. Based on predetermined criteria, it is determined that the Kepler Night Nurse beats the existing solution by 50% to sometimes 99%.

During this calibration step, a maximum of 1% of the total data is stored.

The Activity Report helps caregivers in the nursing home to gather more insights into the lifestyle and schedule of their clients. By recognizing trends in the behavior of the client in an early stage, improvements or deteriorations can be spotted more quickly. For example, when a client needs to use the restroom more often during the night, this may indicate cystitis.

The content of the Activity Report could also be integrated into existing e-health applications. Therefore, the Kepler Night Nurse can be used in a broader spectrum of care innovations and health technology. The information from the Activity Report can be sent in multiple forms, including JSON and PDF, to the Electronic Health Record.


After it has been determined that the Kepler Night Nurse works much better than the old solution (if present), the further roll-out of the care software is executed. The old solution is phased out on a location by location process, with the Kepler Night Nurse replacing it.