Many telecommunication companies of our time are developing in one of the currently popular areas: artificial intelligence, augmented or virtual reality, robots, IoT. Giants such as Google are not interested in just one topic, but in many, highlighting in some cases a whole department for work in a particular area. In the case of a corporation, the robot is engaged in, Waymo, Deepmind - artificial intelligence and neural networks. There are other subsidiaries, and many.
The corporation in some cases creates hybrid projects that use developments from different fields. For example, in the new Google project, the achievements of science and technology from such areas as artificial intelligence, augmented reality, optics, machine learning, and medicine are applied. The company decided to use all this to create a microscope with augmented reality. The system, according to the developers, helps identify cancer diseases. Not all, but with high speed and accuracy.
The results of their work were presented by a team of researchers at a meeting of the
American Association for Cancer Research (
American Association for Cancer Research ). Meetings are held annually, they share experience of the largest experts in this field, who have something to tell, although they also do not mind listening. The researchers published an article on the subject of the speech, telling in detail about the creation of an
“augmented reality microscope” .
According to scientists, their development will help doctors quickly and accurately diagnose the presence of cancer. Another project role is to stimulate the adoption of new technologies by traditional scientific communities. In this case we are talking about deep machine learning.
The system in question consists of a modified light microscope, which allows you to analyze the visible image in real time. And it is not necessary to create a new microscope - you can modify existing solutions. The design is assembled from relatively inexpensive components that can serve for quite a long time. It does not need to buy expensive digital devices that cost research and medical organizations very expensive.
A microscope not only enlarges what is in its field of view, but also helps a person determine what exactly is in the eyepiece. As far as can be understood, this object identification system is based on
TensorFlow . By teaching a neural network (it is part of the entire system), a very effective hybrid of a conventional microscope and a digital system can be obtained. And the work is done in real time.
The general scheme of the system and its picture. The design element of the microscope is a digital camera with a field of view like that of a human eye. The image is transmitted to the computer, where it is processed using a neural network. The results are displayed in the eyepiece, using a specialized AR-screen. As a result, it seems that a person looks at the object itself with ready-made explanations of what exactly fell into the hands of the researcher.The system highlights important elements of the image with arrows, contour, text, there is also a visual feedback. At the moment, the microscope works with two cancer detection algorithms: the first detects metastases in breast cancer, the second is trying to detect prostate cancer. These models operate at a magnification of 4-40 times. If a problem is identified, a microscope leads around the problem area with a green outline.
The accuracy of cancer detection is quite high - 98% in the case of breast cancer and 96% in the detection of prostate cancer. This is higher than with traditional analysis.
According to representatives of Google, all this can be useful to doctors of any country. Now the detection of cancer is not an easy task, but the software and hardware platform allows it to be solved without any problems. If technology is adopted around the world, the number of successfully detected cancers can increase at an early stage, which will increase the life expectancy of people.