Convolutional neural networks trained to take blood for analysis quickly and painlessly

Researchers from the Skolkovo Institute of Science and Technology have developed the first prototype of a medical imaging system. It will allow you to more accurately and less painfully take blood from patients for analysis.

Skoltech’s new development is based on the use of neural networks to analyze near-infrared vein images and project the vein pattern directly onto the patient’s body. The new method will help to simplify venous blood collection and reduce patient discomfort. The revitalized blood collection process will be especially beneficial for people with diabetes.

Details of the study were published in the proceedings of the XVI International Conference on Control, Automation, Robotics, and Technical Vision (ICARCV).

About 20 million blood tests are performed worldwide every day. The authors of the article cite data according to which, in 45% of cases, blood sampling causes some discomfort in patients due to various reasons that make it difficult to access the veins. If the veins are poorly visible and not palpable, there is a risk of injury when taking blood. Also, multiple or inaccurate punctures lead to the risk of tachycardia.

Dmitry Dylov, associate professor at the Skoltech Center for Scientific and Engineering Computational Technologies for Large Data Problems (CDISE), head of the Skoltech Computational Visualization Group, and his colleagues have developed an intelligent near-infrared vein scanner. It allows you to accurately determine the contours of the veins in the legs and arms. It should be noted that this can be done in a fully automatic mode without using any user data.

“Infrared vein scanners are already widely used in clinical practice, but our device is the first development based entirely on advanced artificial intelligence methods. One neural network is responsible for noise reduction and processing of the infrared signal, the second determines the contours of the veins, and the third constantly monitors that the calculated outlines of the vessels coincide with their actual boundaries, ”explains Dylov.

A neural network is a mathematical model, as well as its software or hardware implementation, built on the principle of the organization and functioning of biological neural networks – networks of nerve cells of a living organism. Convolutional neural networks, in turn, aim at efficient pattern recognition, which is part of deep learning technologies. It uses some of the features of the visual cortex, in which the so-called simple cells were discovered that respond to straight lines at different angles, and complex cells, the reaction of which is associated with the activation of a certain set of simple cells. Thus, the idea behind convolutional neural networks is to alternate between convolutional and downsampling layers. The structure of the network is unidirectional (without feedbacks), fundamentally multilayer. For training, standard methods are used, most often the backpropagation method. Function of activation of neurons (transfer function) – any, at the choice of the researcher.

In the new work, scientists have adapted neural networks to create a system that finds a patient’s blood vessels.

At the training stage, the specialists explained to the system “what is good and what is bad”. The neural networks did the rest themselves: they automatically found the optimal settings for new patients, determined the external conditions and even tracked distortions, including those that the system had never encountered before.

Scientists tested the operation of this system on a set of pictures of patients’ hands, and then created a prototype of the device and tested it on volunteers.

The authors of the development note that the scanner can be scaled for use on other parts of the body, as well as used in veterinary clinics to perform complex vein punctures in animals.

If you have found a spelling error, please, notify us by selecting that text and pressing Ctrl+Enter.

Author: John Kessler
Graduated From the Massachusetts Institute of Technology. Previously, worked in various little-known media. Currently is an expert, editor and developer of Free News.
Function: Director

Spelling error report

The following text will be sent to our editors:

130 number 0.276095 time