- Computer Vision
- Image Recognition
- Neural Networks
Our client from the healthcare industry discovered that many patients buy medicines but forget to take them or break the rules of intake. It's bad both for the treatment process and patients' wellness. At the same time, medicine intake is very hard to manage both for medical staff and patients, as well. So, he decided to create a tool for pill intake tracking which can remind patients and notify doctors.
There is a percentage of patients who get subscriptions, buy medicines but never take them or stop taking them in the middle. Our customers care about the patients and want to remind them when they forgot to take a pill in time. So, UKAD created a tool for intake monitoring which can easily notify users when they miss a pill. There were several ways to solve it. The easiest way was to create an app with only two features: manually check taken meds and notify if the check missed. But we decided to act smarter and involve cutting-edge technologies for better intake management. While everyone has a quality camera in the smartphone, UKAD experts decided to utilize computer vision for recognition and fixation of changes in the blisters.
UKAD designed and developed a mobile application because it's the easiest way to stay connected with patients 24/7 in different ways. A possibility to control the intake of medicines provided by a machine learning algorithm that analyzes blisters to find missed pills. If a patient doesn't scan the blister in time or forget to take a pill, the app sends a notification. If nothing happens after notification, the patient receives a call. Such an approach provides monitoring quality just similar to the distribution of medicines in the hospital. A solution powered with YOLO real-time object detection which works extremely fast. We decided to use it to provide an outstanding user experience on most of all devices and make the application as comfortable as possible.
UKAD successfully developed a prototype of a mobile application and a neural network to power it. We also trained our computer vision system with several blister types and got a huge recognition rate. As soon as testing with real medicine was successfully passed, the team and the customer decided that the application is stable and safe to use with real patients.
Frontend and Backend
- .NET CORE 2.0
- CNN Darknet