System of assessment of a status of cervical arteries

The system of assessment of a status of cervical arteries is developed based on algorithms of machine learning and allows to carry out automatic search of cholesteric plaques in arteries in the field of cervical department of results of ultrasonic scanning.

2018: Development

The WaveAccess company in March, 2018 announced development by the independent division of Health Samurai of the module based on algorithms of machine learning for automatic search of cholesteric plaques in cervical arteries based on ultrasonic scanning.

The solution for assessment of a status of cervical arteries includes the portable scanner and service for detailed interpretation of results of scanning by experienced technicians. The neuronet helps to define areas of finding of arteries and to distinguish cholesteric plaques.

After training a system learned to distinguish normal "noise" by videos from anomaly and to find the necessary personnel with a high percent of accuracy. The module divides record into the separate personnel and offers technology of 5 most qualitative pictures with the most successful area for measurement of diameter of a wall of an artery. Also a system offers 20 pictures of an artery where the areas which are of interest are selected. With a task of interpretation of record for measurement of diameter of a wall of an artery algorithms cope with accuracy about 95%, with a task of search of the personnel with potential anomalies — with an accuracy of 80%, told in WaveAccess.

Specialists of Health Samurai taught to distinguish a neuronet normal "noise" by videos from anomaly

The project is custom-made the American medical company, the supplier of the high-tech solution for assessment of a status of arteries.

Solution development demanded close cooperation of our algoritmist and developers with doctors. At the first stage of implementation medical specialists "doobuchal" the solution: browsed the personnel manually, pointed out anomalies, estimated quality accepted by solution module and told what is not enough for them in data. Data prepared manually that required special attention of specialists on both sides — Alexander Azarov, the senior vice president for software development in WaveAccess told.

Earlier interpretation specialists browsed all received video of average 5 minutes, a frame off-screen checking an artery status to reveal pictures, representative for the doctor, and to define on them existence of cholesteric plaques. Measurement of thickness of a wall of an artery and its diameter, determination of a form, the sizes and arrangements of plaques were also performed manually. Preparation of one conclusion could take several hours.

Standard software solutions, according to doctors, did not allow to automate process of "reading" scans — with interpretation only the qualified specialist could cope with medical education. With growth of number of users the supplier of the solution had to increase the staff of highly paid technicians, also the service price grew. At the same time the speed and quality of service did not change. To make a system competitive, it was required to optimize processes: accelerate work of the current staff of specialists and at the same time to increase the accuracy of interpretations.

It was succeeded to solve a problem thanks to the module based on algorithms of machine learning developed by Health Samurai. On a response of the customer, development of Health Samurai allowed to increase the speed of interpretations by 5 times: now it is not required to browse all record entirely — the personnel, necessary for the conclusion, are provided automatically. Besides, the probability of the error connected with a human factor decreased. In general process automation exempted medical employees from routine transactions, and saved the company from need to increase staff of highly paid specialists.