An Automatic Medical Device to Diagnose Safely
Critical Patients in Emergency Departments
The Capillary Refill Time (CRT) is used to monitor the microcirculation of patients suspected of circulatory failure which presents poor tissue perfusion. CRT is a sensitive parameter that has been the subject of several medical studies. These studies are demonstrating that CRT is the right parameter to detect circulatory failure. Until now, CRT is measured manually by practitioners/doctors and is not repeatable and reproducible. The alternative of this visual and subjective CRT evaluation is a blood test of lactate concentration which is an indication to detect circulatory failure. However, such tests take about 20 – 30 minutes to be analysed and is less significant. Every year about 140 000 patients suffer a critical circulatory failure in France and about 40 000 even die. With the use of the DiCART medical device, 10% of these people can be saved.
DIATOMIC has helped our project by providing us guidance during the Periodic Coaches session and providing help when requested and connexion to their network for next steps.
The current solution to detect circulatory failure is either the lactate concentration blood test or performing CRT manually. The lactate concentration is a slow and invasive solution compared to the manual CRT which is a good solution but is operator-dependant as the doctor applies pressure to the tissue. With DiCART, this measure is repeatable, accurate and reproducible, since DiCART is using a high definition image and real time calculation providing a value which allows clinicians to know the CRT result immediately. In terms of costs the blood test to measure the lactate
costs over 10€ whereas DiCART costs only 3€ by patient by 24h as this only the cost of the protector tips used in the device in order to fulfil hygienic concerns and to not have cross-contamination.
The most important result is the first calculation of the CRT with the algorithm we created. It is the first time that CRT is calculated using video imaging and an algorithm, with results comparable to manual CRT.
The current algorithm gives good results, but we are working on an improved new algorithm based on deep learning approach that will further enhance our work. We learned during testing that utilising the physical device, the video treatment needs higher computing power, mostly due to the videos in High Definition and thus in order to embed the algorithm with faster response time, we needed to decrease a bit the quality of video. However, even with this reduction of video quality the CRT calculation is not impacted and an accurate result is displayed in the LCD display.