Diagnosis of the disease is no less important than its treatment. But often due to various reasons it can be difficult a number of additional factors. And if not particularly life-threatening disease, the delay will not give major complications, diseases like cancer requiring rapid diagnosis. And to improve the detection of diseases Google propose to use machine learning algorithms.
To help Google employees decided that according to statistics, the questions of whether the patients had some forms of breast and prostate glands, doctors opinions can coincide only at 48%. To resolve this problem, and it was suggested to use «AI». To create the program, researchers used images of bodies affected by cancer, provided the medical center of the University of Nijmegen. For these purposes a good fit already existing algorithm called Inception (or GoogLeNet).
The training system «feed» a large number of heat maps, i.e., images showing which parts of the tissue can be affected. After some modification the algorithm prepared with the help of heat maps has been improved so that the accuracy reached 89%, compared to the original images that contain a large number «noise». Further it was decided to compare the results and conclusions of the Inception reports, which would give doctors. In the end, after the analysis of 130 of the pictures it became clear that Inception is 16% more effective person specifies the affected areas.
Despite the good results, the authors of the algorithm claim that the model is still far from perfect. It pays no attention to the presence of comorbidities, which can indirectly speak about the presence of the disease. But even in this case, if to complement the work programme, the diagnostician, it can give very positive results and increase the number of identified diseases that could be missed.
One of the main problems of transplantation is that organs intended for transplantation, very few «live» outside of the human body.
according to the Agency «RIA» with reference to the press service of the Novosibirsk state University (NSU), Novosibirsk a large group of scientists, including experts from the Institute of Cytology and genetics, Siberian branch RAS and Novosibirsk state University, has developed innovative technology to create artificial blood vessels based on the cell structure of the human myocardium.
Tiny nanobots, development and improvement of which lately is particularly active, have great potential in medical practice: from targeted drug delivery to diagnostics of diseases, destruction of blood clots and plaque, and even operations.