Artificial intelligence in radiology

The main articles are:

• Radiology
• Artificial Intelligence in Medicine

Advantages of using AI in radiology

Advantages from the use of a neural network in the formulation of radiological diagnoses, which are noted by Russian specialists for August 2019:

• AI allows you to describe an X-ray in 3 seconds - a specialist describes a picture of up to 20 minutes. In the "AI-doctor" bundle, the description of the study is less than 3 seconds. According to current standards, conducting and describing one study can take up to 90 minutes.

• Low system cost

• AI allows you to draw up a second opinion for a radiologist. The "second opinion" allows you to find out the opinion of an independent expert, get more information about the disease and the treatment plan. The reliability of diagnostics in this case increases by 48%.

• AI also allows you to draw up a second opinion for a clinician

• AI helps to separate patient flow and prioritization by removing a significant part of this burden from the doctor

• An equally important point is the possibility of quality control through technology and audit.

• The accuracy of the system in the description of the picture in conjunction with the doctor is 95-98%. The neural network identifies the specific area on which the pathology was found, which allows the doctor to conclude on the basis of the picture very quickly.

How neural networks are trained to describe X-rays

One of the main tasks of AI is to help doctors in preventing medical errors, as well as to bring examinations to a fundamentally new qualitative level due to the accuracy of data analysis and description.

The training of the neural network in radiology is as follows: before starting the test use of the program, more than 200,000 X-rays were analyzed and work continues. The system calculates the error, and then the specialists configure the network. Experienced radiologists with long experience take part in this event. Each X-ray image itself is not only marked, but also analyzed by 3 experts, independently of each other. If the results of the studies coincide with all radiologists, then they are subsequently used by the neural network. Then the trained neural network is connected to radiological systems. After receiving the patient's snapshot, AI, based on previously obtained and corrected research data, makes a conclusion. The doctor, in turn, corrects and supplements the conclusion if necessary.

According to Grigory Reutberg, President of the Clinic of Medicine JSC, expressed in August 2019, a similar system over the next few years will replace 70-80% of radiologists in modern clinics.

Artificial intelligence in cancer diagnosis

Main article: Cancer diagnosis

Artificial intelligence is actively used in research on the development of methods for diagnosing cancer.

2024

A new AI service for the diagnosis of cholelithiasis on computed tomography of the abdominal organs began to work in Moscow

In the capital, for the first time, an artificial intelligence service was launched to detect cholelithiasis on computed tomography. The neural network will automate the detection of gallstones, measure their size and speed up the diagnosis of pathology. This will help increase the chances of successful treatment and recovery of patients. Anastasia Rakova, Deputy Mayor of Moscow for Social Development, spoke about this on November 1, 2024. Read more here.

Domestic AI service surpassed foreign ones in terms of accuracy of detecting nodes in the lungs

The first AI service has been launched in Moscow to identify and assess the displacement of vertebral bodies

A service with artificial intelligence (AI) has become available to Moscow doctors, which helps to identify and assess the displacement of vertebral bodies on radiographs. This was announced to the medical portal Zdrav.Expert on September 9, 2024 by representatives of the Center for Diagnostics and Telemedicine of the Moscow Department of Health. According to them, "this is a new direction in the development of such technologies." In total, more than 50 AI services are available to radiologists in the capital, who find signs of 38 pathologies on radiation studies, said Anastasia Rakova, Deputy Mayor of Moscow for Social Development. Read more here.

Artificial intelligence trained to find CNS defects in fetus during ultrasound

On August 19, 2024, Yandex announced the development of a neural network that will help doctors detect the symptoms of spina bifida, a severe congenital disease of the central nervous system in children, during an ultrasound study of pregnant women. This congenital pathology is difficult to diagnose as it occurs once per thousand newborns. It often leads to severe disability. With the help of technology, medical specialists will be able to see signs of this disease at an earlier date and send the patient for additional examination. The solution is available free of charge to all doctors and medical experts on the website of the Spin Bifida Foundation, which initiated the first such project in Russia. Read more here.

Four more regions connected to the Moscow platform with AI services

Four more regions joined the Moscow platform "MosMedII." These are Kaliningrad, Lipetsk and Chelyabinsk regions, as well as the Yamalo-Nenets Autonomous Okrug. Now doctors of six constituent entities of the Russian Federation can use artificial intelligence (AI) services to process radiation studies. This was reported on August 13, 2024 at the Center for Diagnostics and Telemedicine with reference to the words of Anastasia Rakova, Deputy Mayor of Moscow for Social Development. Read more here.

An AI service has been created in Russia to automatically assess heart research

11 regions of the Russian Federation connected to the Moscow digital platform "MosMedII" for processing CT and X-ray images

In July 2024, 11 regions of Russia began connecting and testing the Moscow digital platform MosMedII, designed to process computer tomograms and X-rays using artificial intelligence (AI). Read more here.

How neural networks help doctors in Moscow: More than 12 million images have been processed, signs of 37 different diseases are detected

The capital has been using computer vision technologies in medicine for the fifth year. During this time, neural networks processed more than 12 million images and learned to detect lung disease, heart disease and signs of oncology. Artificial intelligence (AI) analyzes fluorograms, mammograms, radiographs, magnetic resonance and computer tomograms, finds signs of pathologies and takes the necessary measurements. On May 31, 2024, the Moscow Department of Information Technology and the Moscow Department of Health told about how smart algorithms help doctors. Read more here.

The Moscow library of datacets for AI has been replenished with new data sets

The Moscow library of data sets for AI has been replenished with new data sets. This was announced on May 3, 2024 by the press service of the Center for Diagnostics and Telemedicine DZM. Read more here.

NtechLab registers product to diagnose ischemic and hemorrhagic strokes

The company's medical product - NtechMed CT Brain - has received a registration certificate and is ready for implementation in public and private clinics. The company previously announced its entry into the medical technology market. NtechLab (Ntech Lab) announced this on April 26, 2024. Read more here.

SberMedII registered AI-service "Brain CT" in Roszdravnadzor

The company SberMedII registered in Roszdravnadzor AI the service "-brain CT " a system for supporting medical decision-making for diagnostics. stroke Algorithm helps the doctor quickly analyze CT images and make the correct diagnosis, which in critical cases allows you to start treating the patient in a timely manner and save his life. This was SberMedII announced on April 17, 2024. More. here

2023

Moscow scientists presented an assessment of the maturity of AI services in radiation diagnostics

Moscow scientists assessed the maturity of artificial intelligence services operating in the capital's healthcare at the end of 2023. For this, a maturity matrix was used, previously developed by specialists from the Center for Diagnostics and Telemedicine of the city Department of Health. Using the service, radiologists can choose proven and reliable solutions based on artificial intelligence (AI). The NPCC DiT DZM announced this on May 14, 2024. Read more here.

In Moscow, for the first time, artificial intelligence began to detect ENT diseases on X-rays

To Moscow neuronets For the first time, they began to recognize the signs of ENT diseases in the x-ray pictures. Smart algorithms are able to detect a serious pathology - sinusitis from the results of sinus radiography. This is the most common study to detect diseases of the nasal sinuses. In 2022 Russia , more than three million such surveys were conducted in. A service based on () artificial intelligence AI will help doctors get preliminary results immediately after the procedure. This was Center for Diagnostics and Telemedicine (NPCC DiT DZM) announced on November 14, 2023. More. here

Moscow doctors can already use 5 neural networks to help in determining diseases of the chest organs

Moscow doctors can already use 5 neural networks to help in determining diseases of the chest organs. This was announced on October 31, 2023 by the Center for Diagnostics and Telemedicine (NPCC DiT DZM). Read more here.

AI has learned to find intracranial neoplasms on magnetic resonance imaging of the brain

Artificial intelligence (AI ) has learned to find intracranial neoplasms on magnetic resonance imaging (MRI) of the brain. Now neural networks in the capital's experiment on the introduction of computer vision technologies into healthcare are already working in 27 clinical areas. This was announced on October 9, 2023 by Anastasia Rakova, Deputy Mayor of Moscow for Social Development.

In an experiment to introduce computer vision into the Moscow health care system, an artificial intelligence service was launched in the next direction. Neural networks now help doctors find intracranial neoplasms when conducting MRI tests of the brain. Thus, Moscow radiologists can use more than 50 algorithms in 27 areas. It is important that this service was created by a Russian company. Domestic developers are actively involved in our experiment and, accordingly, offer more and more high-quality solutions for the market. This year alone, nine manufacturing companies joined us, seven of which are Russian, - said Anastasia Rakova.

The use of computer vision technologies in medicine makes it possible to reduce the diagnostic time. In total, with the help of neural networks, doctors analyzed more than 10.5 million images.

Timely diagnosis is very important. Magnetic resonance imaging is the main way to detect neoplasms in the brain. The creation of artificial intelligence services in this direction is one of the difficult tasks, which has many restrictions, so there are not many developments here for October 2023. It is all the more valuable that now doctors specializing in working with MRI studies also have the opportunity to use neural networks in the diagnosis of such a serious pathology, - said Yuri Vasiliev, chief radiologist of Moscow, director of the Center for Diagnostics and Telemedicine of the Moscow Department of Health.

Algorithms mark areas of possible tumour formations with color hints on a tomogram. The neural network automatically makes measurements that are important to the doctor at diagnosis and compiles a description.

The fifth comprehensive AI service in Moscow can reveal six pathologies at once in one picture

Artificial intelligence processed 1 million mammograms of metropolitan patients in 3.5 years

The fourth complex of AI service for the diagnosis of diseases has started in Moscow

The fourth comprehensive artificial intelligence service has appeared in Moscow, which can simultaneously detect several diseases on a computed tomogram of the chest organs. The algorithms work within the framework of the Moscow experiment on the use of computer vision technologies in radiation diagnostics. This is the fourth comprehensive solution in the domestic market of algorithm developers. This was announced on September 27, 2023 by the acting Deputy Mayor of Moscow for Social Development Anastasia Rakova. Read more here.

Now artificial intelligence helps Moscow doctors identify signs of fractured ribs on CT

In the Moscow Art Intelligence, on CT studies of the chest organs, he learned to determine the fracture of the ribs. This tool will help doctors in describing complex diagnostic cases, for example, in combined pathology or severe injuries. This direction will be included in the complex service of artificial intelligence, that is, algorithms will be able to determine in one picture the signs of many pathologies at once. This was announced on August 22, 2023 by the Deputy Mayor of Moscow for Social Development Anastasia Rakova. Read more here.

In Moscow, the official library of data sets for the healthcare sector has been replenished by 50%

In Moscow, the official library of data sets for the healthcare sector has been replenished by 50% per year.

Ilya Tyrov, deputy head of the Moscow City Health Department, spoke about this on August 21, 2023. Read more here.

Moscow scientists have developed principles for standardizing datacets for teaching AI in medicine

Moscow scientists have developed principles for standardizing datacets for training artificial intelligence in medicine.

The developers train neural networks to independently find pathologies in radiation studies of patients using datacets - sets of impersonal radiological images with signs of certain diseases. This was announced on July 5, 2023 by Anastasia Rakova, Deputy Mayor of Moscow for Social Development. Read more here.

Revolution in radiotherapy. New AI system created to precisely kill cancer cells

At the end of June 2023, British researchers from Addenbrooke's Hospital announced the development of a new technology based on artificial intelligence to accurately destroy cancer cells. Read more here.

Comprehensive AI service in Moscow can now simultaneously detect up to 10 pathologies on a computed tomogram

Moscow has expanded the capabilities of a comprehensive artificial intelligence service. The neural network can now simultaneously detect up to 10 pathologies on a computed tomogram. The diseases that algorithms help doctors determine on CT images of the chest included adrenal glands. The Center for Diagnostics and Telemedicine announced this on June 27, 2023. Read more here.

The second comprehensive artificial intelligence service for diagnosing 7 diseases at once has been launched in Moscow

A To Moscow second comprehensive service has appeared in, artificial intelligence which computer can simultaneously detect seven diseases on a tomogram. In one study of the chest neuronet , it is able to recognize the signs, lung cancer COVID-19 osteoporosis spine, aortic aneurysms, pulmonary hypertension hydrothorax, ischemic heart disease by the degree of calcification of the coronary arteries. This is the second comprehensive solution from domestic developers - doctors will be able to choose which service is convenient to use. Anastasia Rakova, Deputy Mayor of Moscow for Social Development, spoke about this on June 8, 2023. Read more here.

Artificial intelligence now helps Moscow doctors diagnose multiple sclerosis

Artificial intelligence analyzed more than 27 thousand medical images of patients with Yamalo-Nenets Autonomous Okrug

More than 27 thousand studies conducted by patients of the Yamalo-Nenets Autonomous Okrug (Yamalo-Nenets Autonomous Okrug) analyzed neural networks developed in the capital. This was announced on May 4, 2023 by Anastasia Rakova, Deputy Mayor of Moscow for Social Development. Read more here.

AI learned to identify signs of ischemic stroke on brain CT scan

In Moscow, with the help of AI, more than 9 million radiation studies were analyzed. 44 AI services are available in 19 directions

Moscow summed up the results of three years of experiment on the introduction of computer vision in radiation diagnostics. As of mid-March 2023, 44 artificial intelligence services in 19 areas are available to radiologists in the capital. Neural networks have already analyzed more than 9 million radiation studies, Zdrav.Expert was told on March 15, 2023 at the Center for Diagnostics and Telemedicine of DZM. Such technologies accelerate patient outcomes and save physicians time describing medical images.

As the Deputy Mayor of Moscow for Social Development Anastasia Rakova said, in 2020 an unprecedented experiment was launched in Moscow to introduce computer vision into radiation diagnostics. Based on automation and a scientific approach, new ways of organizing medical care are being formed, aimed at increasing its availability and quality.

"So, in 2020, we created a reference center where radiologists describe more than 80 thousand CT scans, MRI scans, mammograms and X-rays around the clock. Such a model of radiation diagnostics increased labor productivity by half and significantly reduced the time for preparing the conclusion, - said the deputy mayor.

According to Anastasia Rakova, as a result of the experiment, scientific methodologies were developed and introduced, on the basis of which more than 200 reference data sets were prepared, the first open library of data sets for the healthcare sector in Russia was created. Based on the scientific results of the experiment, 11 national standards in the field of artificial intelligence in healthcare were developed and approved, 10 of which entered into force. In 2022, the Yamalo-Nenets Autonomous Okrug also joined the testing of artificial intelligence services, "the deputy mayor said.

The experiment is carried out by the Moscow Department of Health on the basis of the Moscow Center for Diagnostics and Telemedicine DZM together with the Department of Information Technologies. It involves more than 150 medical institutions, including children's, as well as about 20 IT companies - developers of artificial intelligence, more than 1200 units of diagnostic equipment are involved. More than 20 algorithms participating in the project received state registration as a medical device, which means that they can work outside the framework of the experiment.

"The experiment aims to create a reliable assistant for the doctor. We annually collect feedback from colleagues using artificial intelligence and transmit this information to developers in order to constantly improve it. Over the past year, the number of doctors in Moscow who actively use artificial intelligence services in everyday practice has grown 1.5 times. At the same time, specialists began to positively assess the work of algorithms by 20% more often, "said Yuri Vasiliev, chief freelance specialist in radiation and instrumental diagnostics in Moscow, director of the Moscow Center for Diagnostics and Telemedicine of the City Health Department.

In 2023, the project will include comprehensive solutions for various anatomical areas and modalities. As of mid-March 2023, Moscow doctors have access to one such service for analyzing CT examinations of the chest organs, it finds signs of 9 pathologies. It is also planned to automate many more routine measurements. In addition, in the Center, developers can conduct clinical and technical tests of their service in order to obtain a registration certificate and become a full-fledged medical device.

Since 2020, Moscow has been actively introducing technologies based on artificial intelligence into the capital's healthcare. The work is being carried out as part of a Moscow experiment on the use of computer vision technologies. Thanks to the automation of routine processes, doctors have more time to analyze the patient's condition. Artificial intelligence allows detecting signs of dangerous diseases on the images of radiation studies, which the patient does not suspect, including, lung cancer COVID-19 osteoporosis spine, aneurysms aortas ischemic disease,,, hearts stroke pulmonary hypertension hydrothorax, as well as hernia breast cancer of the spine, and other flat-footedness diseases. At the same time, work is underway to select the best artificial intelligence services for radiation diagnostics doctors, training of medical staff to work with neural networks is being carried out, and the expansion of opportunities for the introduction of smart services continues.

In 19 areas, the developers went to streaming research, testing and refinement of models are being carried out for the rest. It is important that it is conducted on the basis of a real flow of research, and doctors constantly provide feedback on the work of algorithms. Equal conditions have also been created for all participants: the catalog of solutions based on artificial intelligence is constantly updated, a leadership board of services is compiled every month, the DZM Center for Diagnostics and Telemedicine told.

Since 2023, Moscow was the first in the country to introduce a special tariff under the compulsory medical insurance to analyze the results of preventive mammographic examinations using artificial intelligence systems.

Moscow has expanded the capabilities of a comprehensive artificial intelligence service

Moscow scientists have developed a methodology for preparing datacets for testing AI services for the analysis of radiation research

Moscow specialists have created a methodology for preparing medical datasets for testing neural networks for the analysis of radiation research. This was announced on February 16, 2023 by the NPCC DiT DZM. With the help of impersonal radiological images with signs of diseases, smart algorithms learn to independently find pathologies. The recommendations are intended for doctors of any specialty who organize and perform markup of medical datasets. The prepared approaches will make it possible to universally unify the development of datacets, ensure their quality, as well as accelerate the introduction of artificial intelligence in medicine to analyze patient research.

The safety and quality of neural networks is directly due to data sets, so the process of their formation requires an understandable methodology for specialists. Moscow guidelines are based on extensive practical experience in the formation of patient research kits for testing smart algorithms. The methodology describes universal approaches and can be used by doctors of various specialties in any field health care where the introduction of neural networks is relevant. The diagnostic center telemedicine Department of Health and has a large expertise in the field medicine information technology and specialists test the introduction of artificial intelligence into radiation diagnostics, develop draft rules and standards for its use in specific clinical scenarios.

These guidelines were the result of combining world practices and the own experience of the Department of Health's Center for Diagnostics and Telemedicine on the implementation of artificial intelligence services.

The methodology was developed and tested during the Moscow experiment on the use of technologies in the field of computer vision for the analysis of medical images and research work "Scientific justification of the methodology for the use and methods of assessing quality (artificial intelligence) in diagnostics." It contains a description of practical approaches in the planning and creation of datasets necessary for the testing and application of artificial intelligence technologies in healthcare.

The Moscow experiment is a scientific study of medical artificial intelligence. As of February 2023, in an experiment of more than 40 services in 19 clinical areas, in almost three years, neural networks have already analyzed more than 8.5 million images obtained using radiation methods of research of patients of medical institutions in the capital.

Moscow has been engaged in the digitalization of the health care system for more than 10 years. The basis of this process is a single digital platform. Including thanks to her, the complex of social development of Moscow, together with the city Department of Information Technologies, began an experiment on the introduction of computer vision technologies in medicine on the basis of the Center for Diagnostics and Telemedicine of the Moscow Department of Health. It has become a platform for the development of artificial intelligence technologies in radiation diagnostics, as well as to support domestic developers.

Tests of a domestic AI service for analyzing photographs of the fundus and OCT scans have been completed

On the basis of the Federal Scientific and Clinical Center of the FMBA of Russia, a clinical trial of the Retina.AI cloud platform was completed. The domestic service for ophthalmologists based on artificial intelligence is aimed at preventing blindness and low vision in people suffering from diabetes mellitus and age-related retinal changes. The FNCC FMBA of Russia announced this on February 8, 2023. Read more here.

2022

Comprehensive AI service will help identify signs of coronary heart disease on CT scans

To Moscow In, they have expanded the capabilities of a comprehensive service - artificial intelligence neuronets they will detect signs of paracardial fat on CT scans, which will help in the diagnosis warmly of vascular diseases. Thus, now a smart assistant will help determine the signs of eight pathologies at once. This was announced on December 8, 2022 by the deputy head of the capital. Department of Health Ilya Tyrov More. here

In Moscow, 50% of radiation studies of polyclinic patients are processed by artificial intelligence

To Moscow At more than two years neuronets , doctors help. to radiologists Their capabilities are regularly expanded and the proportion of images that are processed by the smart is constantly algorithms growing. At the beginning of December 2022, already half of all radiation studies of patients policlinics Moscow analyze services. This was artificial intelligence announced on December 1, 2022 by the deputy head of City Health Department Ilya Tyrov. More. here

HUB Telemed platform allows YNAO radiologists to work with 7 AI services

More than 4 thousand radiation studies of patients of the Yamalo-Nenets Autonomous Okrug have been analyzed since the start of the launch of a pilot project of the Moscow Health Department to connect medical organizations in the region to the Moscow experiment. As of November 2022, the Moscow HUB Telemed platform allows YNAO radiologists to work with 7 artificial intelligence services in 4 areas of research. This was announced on November 29, 2022 by the Center for Diagnostics and Telemedicine (NPCC DiT DZM). [2]Подробнее #.2A_.D0.90.D0.BD.D0.B0.D0.BB.D0.B8.D0.B7_.D0.B1.D0.BE.D0.BB.D0.B5.D0.B5_4_.D1.82.D1.8B.D1.81.D1.8F.D1.87_.D0.BB.D1.83.D1.87.D0.B5.D0.B2.D1.8B.D1.85_.D0.B8.D1.81.D1.81.D0.BB.D0.B5.D0.B4.D0.BE.D0.B2.D0.B0.D0.BD.D0.B8.D0.B9_.D0.BF.D0.B0.D1.86.D0.B8.D0.B5.D0.BD.D1.82.D0.BE.D0.B2 здесь.

Radiologists in Moscow began to look for pathologies of the spine and adrenal glands using artificial intelligence

Neural networks in Moscow have learned to determine the signs of two more types of diseases. Now artificial intelligence is able to diagnose protrusion and herniation of intervertebral discs on MRI, as well as an adrenal gland tumor on CT. Smart algorithms help radiologists in 17 areas of research. This was announced on November 11, 2022 by the deputy head of the Moscow City Health Department Ilya Tyrov. Read more here.

The number of doctors in Moscow who actively use artificial intelligence services in their work has grown 1.5 times over the year

The number of doctors in Moscow who actively use artificial intelligence services in their work has grown 1.5 times over the year. This was announced on October 27, 2022 by the Center for Diagnostics and Telemedicine DZM. At the same time, specialists began to positively assess the work of neural networks by 20% more often than in 2021, said Ilya Tyrov, deputy head of the Moscow City Health Department.

Artificial intelligence services of Moscow health care are actively developing: their accuracy is growing, they are learning to identify new pathologies. For example, as of October 2022, they are already used in 15 different areas. In addition, we continue to improve the services already implemented and analyze feedback from specialists. In the future, we will adjust the work of artificial intelligence taking into account the information received, - said Ilya Tyrov.

Moscow specialists have changed the approach to assessing artificial intelligence services that work as part of an experiment to introduce computer vision into medicine. Now the clinical assessment set by doctors, along with the assessment of technical parameters, will have an impact on the participation of a particular neural network in the experiment, and, of course, on determining the best solution for future implementation in medical organizations.

Clinical evaluation becomes a priority. It is formed from indicators of accuracy of visual marking - delineation, and agreement with the conclusion - to what extent the conclusion of the neural network, according to the doctor, accurately determines and interprets signs of pathological changes in the images. Our specialists conclude: complete compliance, incorrect assessment, false positive or false negative result of artificial intelligence. Such work will increase the quality characteristics of artificial intelligence services, neural networks will become more valuable for radiologists, "said Yuri Vasiliev, director of the Center for Diagnostics and Telemedicine.

Since the beginning of the Experiment on the introduction computer of vision into practical health care neural networks, more than 7 million studies have already been processed. Algorithms allow you not to miss even the minimum deviation in the radiation study. Artificial intelligence services for automatic analysis of medical images are being tested To Moscow in 2020. After successfully checking the diagnostic and functional performance indicators for compliance with the established standards, they are connected to, Unified Radiological Information Service EMIAS uniting all departments of the Moscow radiation diagnostics. As of clinics and hospitals October 2022, there are already about 150 such medical institutions. Doctors can use artificial intelligence to identify signs of disease. Algorithms are tested and introduced into practical medicine by specialists from the Moscow Center for Diagnostics and. telemedicine

The experiment on the introduction of computer vision into medicine was launched jointly by the Moscow Social Development Complex on the basis of the Moscow Center for Diagnostics and Telemedicine with the support of the city Department of Information Technologies. The center has become a platform for the development of artificial intelligence technologies in Russia.

Moscow opened a digital library for neural networks on the main modalities of radiation diagnostics

On October 18, 2022, the Center for Diagnostics and Telemedicine DZM informed Zdrav.Expert that a digital library of impersonal data sets for assessing and training neural networks was expanded in Moscow. All of them are published on the platform of the NPCC DiT DZM: MosMedII (Mosmed.ai). Read more here.

More than a third of medical images of Muscovites are processed by neural networks

In 2022, more than 40% of radiation studies of patients Moscow in medical institutions are processed with. In artificial intelligence 2021, this figure was about 30%. This was announced on September 7, 2022 Zdrav.Expert by representatives with DMS Diagnostic and Telemedicine Center reference to the words of the deputy head. City Health Department Moscow Ilya Tyrov More. here

A domestic analogue of the automated workplace of a radiologist was assembled in Yakutia

The Yakut company "Cyberia" carried out the commissioning of a fully domestic automated workplace (AWS) of a radiologist. This product of the company will be the next step in the development of the company. Earlier, Cyberia developed methods for analyzing medical images using artificial intelligence technologies to quickly and accurately diagnose COVID-19 and strokes. Read more here.

In Russia, developed an AI system for the diagnosis of dental anomalies in 1 second

Russian scientists have developed a neural network that allows you to quickly and with high accuracy diagnose common dental anomalies. This was announced on March 14, 2022 by Zdrav.Expert at NUST MISIS. A distinctive feature of the development is its compatibility with almost any personal computer. Read more here.

2021

Announcement of AI system that detects prostate cancer during routine CT scans

In mid-July 2021, researchers from RMIT University in Melbourne (Australia) created an AI system that will be able to detect prostate cancer during standard computed tomography. Usually, it is difficult to detect prostate cancer in CT images, and pronounced radiation makes CT unsuitable as a screening method. However, if men undergo abdominal or pelvic scans for other reasons, the proposed system would detect prostate cancer and start treatment early. Read more here.

A neural network has been created that calculates the likelihood of developing lung cancer from fluorography and CT

At the end of May 2021, it became known about the creation of a neural network that calculates the risk of malignant neoplasms from computed tomography and fluorography images. Kiran Venkadesh from the Institute of Health Sciences at Radbud University (Netherlands) and his colleagues developed an algorithm that evaluates the malignancy of pulmonary nodules found in screening CT . To train the algorithm, 16,077 images of pulmonary nodules (including 1,249 malignant ones) collected during the National Lung Screening between 2002 and 2004 were used. Pulmonary nodules are unusually dense structures measuring three centimeters or less. Usually these are benign neoplasms, but in 20% of cases such CT changes indicate a malignant tumor.

The neural network developed by Venkadesh and his colleagues could distinguish between both varieties of pulmonary nodules and estimate the probability of developing cancer with an accuracy of 82-93%, not inferior in this regard to 11 recognized experts in this field. This allows you to use this AI system as an electronic assistant to pulmonologists, oncologists and radiologists..

The performance of the algorithm was compared to the Canadian early detection model of lung cancer (PanCan). The researchers found that in the full cohort, the new algorithm significantly outperformed the PanCan model.

This deep learning algorithm will help radiologists optimize lung cancer screening recommendations and ideally reduce unnecessary diagnostic interventions, the authors wrote.

The developers plan to improve the work of the neural network by adding to it the ability to analyze and compare images taken at different times. This, the researchers hope, will significantly increase the accuracy of its work and allow AI to be used as a criterion for the need for repeated examinations.^[1]

Artificial intelligence helps to identify oncology in residents of Karachay-Cherkessia

On May 8, 2022, SberMedAI (SberMedAI) spoke about the experience of using artificial intelligence to early identify signs of cancer. An example was the experience of using the technology at the Karachay-Cherkess Cancer Dispensary named after S.P. Butova. Read more here.

CT in conjunction with machine learning began to be used to predict serious heart problems

In early May 2021, the researchers reported that the use of machine learning in coronary CT angiography makes it possible to better predict the risk of adverse cardiac events, for example, the appearance of unstable angina. Read more here.

AstraZeneca implements a project for retrospective analysis of CT images of the lungs using AI in Nizhny Novgorod and St. Petersburg

Biopharmaceutical The company AstraZeneca"" together with regional scientific and medical institutions is implementing a project for retrospective analysis of computed tomography images () CT of lungs in and. Nizhny Novgorod As St. Petersburg part of the project, CT images taken earlier to diagnose coronavirus the infection were checked with the help of for artificial intelligence presence in the lungs new growths. This became known on August 25, 2021. More. here

ITMO created a method for recognition by a neural network of brain tumors from MRI images

OneCell has developed a comprehensive AI platform for diagnosing cancer

On February 12, 2021, it became known what he medical startup OneCell had developed. comprehensive platform with AI for the diagnosis of cancer

As reported, OneCell is a telemedicine platform with artificial intelligence and diagnostic equipment in oncopathology. Read more here.

2020

Care Mentor AI in Moscow medical organizations determines the degree of COVID-19 damage from CT scans

On October 27, 2020, Care Mentor AI announced that the AI service of the same name, analyzing CT studies and detecting signs of coronavirus infection on them, was connected to the Unified Radiological Information Service (ERIS EMIAS). The Care Mentor AI system determines the percentage and degree of lung damage in patients with COVID-19 and issues imaging of finds on CT lung series. Read more here.

Announcement of Spline.ai - AI systems for processing X-rays

In mid-October 2020, the developer of computer data processing technologies Xilinx and the company Spline.ai, working on the use of artificial intelligence in medicine, presented an open source deep learning algorithm that allows you to process X-rays. Read more here.

Care Mentor AI using Skoltech supercomputer will create a service to determine the degree of COVID-19 damage

As part of the program to combat COVID-19 of the Skoltech Center for Scientific and Engineering Computing Technologies for Tasks with Large Data Arrays (CDISE), Care Mentor AI data scientists were able to use the Zhores supercomputer to improve the accuracy of determining pathologies. This was announced on September 2, 2020 by Care Mentor AI. Read more here.

Announcement of Siemens AI-Rad Companion - AI systems for automating routine tasks in brain and prostate MRI

At the end of August 2020, Siemens Healthineers introduced two new applications for interpreting artificial intelligence (AI) based MRI images. The new software aims to free radiologists from routine tasks during MRI scans of the brain or prostate. Read more here.

Botkin.AI platform helps Moscow doctors detect lung cancer on CT scans

In Moscow, artificial intelligence began to analyze computer tomograms and detect signs of lung cancer on them. This was reported on August 18, 2020 by the Skolkovo Foundation. To do this, the Botkin.AI platform is used, developed by the resident of the Skolkovo Foundation by Intellectodzhik. The service is integrated with the Unified Radiological Information System of Moscow. Read more here.

Sberbank's free AI service will help identify changes in the lungs, including in COVID-19

The SberHealth service and the SberCloud cloud platform, part of the Sberbank ecosystem, launched a joint project on July 14, 2020 to recognize lung computed tomography images. Read more here.

The first service for analyzing X-rays using a neural network has been launched in Moscow

The Unified Radiological Information Service (ERIS), to which the diagnostic equipment of Moscow is connected, had integrated software Care Mentor AI, which analyzes and screens chest X-rays for various pathologies, including such socially significant ones as lung cancer, tuberculosis, pneumonia. This Zdrav.Expert became known on July 3, 2020. Read more here.

CT scanner maker releases AI system that makes diagnoses in 10 seconds

At the end of May 2020, the Chinese manufacturer of computer tomographs Imsight Technology introduced an AI system for CT diagnostics of the lungs, which establishes a preliminary diagnosis in 10 seconds. The development was called Imsight CT Analysis System. Read more here.

2019

Facial recognition software can recognize a person by an MRI image of the brain

At the end of October 2019, it became known that facial recognition software can accurately match people's photos with MRI images of the brain.

Conventional facial recognition software correctly identified volunteers in more than 80% of cases, however, researchers believe that if more patients are recruited, accuracy will decrease. It is assumed, however, that such software functions can threaten patient privacy, because the laws do not have time to change as quickly as technology. The NYU report already warned scientists about the potential risks of using facial recognition software. This technology is increasingly used by the police to spy on and search for suspects, but regulatory regulations are still seriously behind reality.

Researchers at the Mayo Clinic found that thanks to publicly available facial recognition software, they were able to correctly match patients' MRI photos in 83% of cases. One of the researchers, Christopher Schwarz, reported that the Mayo Clinic team decided to conduct such a study, noting the high quality of images used to study the brains of patients with Alzheimer's disease and dementia. Usually, MRI images show the contour of the head, including skin and fat, but do not determine bones and hair, but the software turned out to be enough for this information.

84 volunteers participated in the study, and facial recognition software successfully correlated their MRI images with photographs. Despite the amazing accuracy of the software, researchers are not ready to trust the results so easily. Professor of radiology Eliot Siegel noted that with an increase in the sample size, the accuracy of "recognition" of photographs will surely fall.^[2]

Russian platform for diagnosing cancer using AI attracted 100 million rubles

The Intellect company, which develops the Botkin.AI project, has raised 100 million rubles. from the funds Digital Evolution Ventures (created with the participation of Rosatom) and RBV Capital (founded by R-Pharm Alexei Repik and the Russian venture capital company), as well as current investors - Primer Capital and ExpoCapital. Botkin.AI is developing an artificial intelligence (AI) based system to analyze and determine pathologies in CT, X-ray and mammography images. The company says it has already implemented pilot projects for the early diagnosis of lung cancer in four regions of Russia. Read more here.

Russian clinic introduced a neural network of a domestic developer to describe X-rays

The Russian private clinic "Medicine" has introduced a neural network for describing X-rays, developed by Care Mentor AI (Care Mentor AI). This is a unique project for Russian medicine: instead of the 20 minutes that a doctor spends describing an X-ray, artificial intelligence does this work in just 3 seconds. That is, by the time the patient is dressed after the study, the description of the picture and the preliminary diagnosis will be ready. Read more here.

AI began to predict children's abilities from MRI images

In mid-March 2019, a study was published according to which artificial intelligence is able to predict the abilities of children from MRI images. By analyzing white matter connections in a baby's brain at birth, the AI algorithm can predict their cognitive development levels at age 2 and offer necessary interventions to at-risk children, the researchers say.

Although evidence suggests that the underlying circuitry of the human brain already exists at birth and that the white matter connectome supports the development of brain functions, exactly how this happens is unknown. Therefore, a team of researchers from the University of North Carolina School of Medicine, led by Dr. Jessica Girault, decided to train AI to analyze white matter connections on MRI of the brain at birth in full-term babies.

Using diffusion-weighted MRI images obtained shortly after birth, the researchers allowed AI to allocate 75 full-term infants to scores above or below the average cognitive development level on the ELC scale used at age 2. AI was found to predict with 98% accuracy the correlation between measures of white matter associations and the actual cognitive estimate of these children at 2 years of age. When tested on a separate sample of 37 premature infants, the algorithm demonstrated 96% accuracy.

Thus, the study showed that the white matter network at birth has a high degree of predictability and may be a useful biomarker in imaging. The fact that the researchers were able to replicate the findings in the second group of children is strong evidence of the reality of their discovery. Potentially similar AI can be used to identify violations early and make decisions about corrective measures.^[3]

AI first began to successfully predict metastases in MRI images of the mammary glands

In late January 2019, researchers reported being able to use a relatively small dataset to train artificial intelligence to evaluate diagnostic images. According to an article published in the Journal of Digital Imaging, the research team prepared a program to predict the chances of developing axillary lymph node metastases in breast cancer according to MRI. The AI program has only achieved 84% diagnostic accuracy, but the team is going to collect more data for training the algorithm to eventually use it in clinical practice. Read more here.

2018

Artificial intelligence attracted to ultrasound diagnostics of pregnant women

A new type of fetal testing for pathologies that a doctor is unable to notice has been launched in a British hospital. The system based on artificial intelligence includes 350,000 images classified by one or another deviation^[4] of^[5].

According to Engineer, ultrasound diagnostics with artificial intelligence are called ScanNav and are designed to give the doctor additional information in real time. As a result, AI allows the specialist not to doubt that all views are taken into account. The latter is particularly relevant due to fetal movement in the womb.

So far, the technology is being tested in test mode in obstetrics, but in the future the development is planned to be applied in various fields of medicine. By the way, AI diagnostics have already been given high hopes in Japan, which is experiencing a shortage of doctors, and in China, artificial intelligence was given a medical license altogether.

Artificial intelligence learned to detect Alzheimer's disease 6 years ahead of doctors

Artificial intelligence can diagnose Alzheimer's much earlier than doctors. Scientists trained a computer system to detect the first signs of severe disease, difficult for the eyes, in positron emission tomography (PET) images. The results of the study on November 6, 2018 published the journal Radiology.

To diagnose Alzheimer's disease, a team of researchers at the University of California created a self-learning AI algorithm. To train the system, scientists used more than 2,100 brain images of over 1,000 patients, which were taken using PET scans, reports The Sun.^[6]

Positron emission tomography allows you to assess the metabolic activity of the brain. With Alzheimer's disease, the metabolic rate in certain areas of nervous tissue decreases, and artificial intelligence has learned to detect these barely noticeable changes. 90% of the shots served to train the algorithm, and the remaining 10% were involved during testing.^[7]

During the final trial, the system managed to identify signs of dementia in 100% of cases in the images of 40 patients, and tracked them on average six years earlier than the official diagnosis was made.

We are very happy with the work of the algorithm. It identified all patients who were subsequently diagnosed with Alzheimer's disease, - said one of the authors of the study doctor Dzhaye Huo Son (Jae Ho Sohn).

In turn, Professor Noel Sharkey from the University of Sheffield, UK, said that although the sample size and test program were relatively small, the results obtained are very promising and indicate the feasibility of a larger study.

According to experts, earlier diagnosis of Alzheimer's disease will slow down or even stop its development.

Facebook speeds up MRI scans 10 times thanks to artificial intelligence

In August 2018, Facebook announced a research project aimed at increasing the speed of MRI scanning using artificial intelligence technologies.

Together with the University of New York , an AI system was created, trained on about 3 million MRI images of the heart, liver and bones, as well as on 10 thousand studies with diagnoses. Facebook in this project is responsible for computer vision and imaging technologies.

According to the results of the training, the neural network allowed MRI devices to quickly scan human organs and give results due to the fact that the equipment began to study only some areas of the body. Thanks to the use of the AI algorithm, the speed of medical scanners has increased by up to 10 times.

Using AI, you can collect less data and, as a result, speed up scans, while preserving or even expanding information-rich MRI images. The challenge is to train artificial neural networks to recognize the basic structure of images to fill in the gaps that occur with accelerated scanning, Facebook said in a blog post. - This method is similar to how people process sensory information. When we learn the world, the brain often receives an incomplete picture, as in the case of blurred or dimly lit objects, and subsequently completes it. We need to use this information for medical benefit.

Facebook emphasizes that the training of the AI system took place using anonymous data in full compliance with the HIPAA privacy norms.^[8]

Earlier in 2018, CNBC reported that Facebook was in talks with hospitals about sharing patient data, but the project was closed after its publicity.

Google algorithm for analyzing thoracic fluorography

At the end of March 2018, Google demonstrated artificial intelligence for fast and efficient processing of thoracic fluorography. The development was presented at the EmTech Digital conference in San Francisco.

Google has created a deep learning model that can improve on the basis of a small number of annotated (with manually marked areas where deviations are visible) medical images and allows you to simultaneously identify the disease and highlight it in the image.

To train the algorithm, ChestX-ray8 was used - the largest open database of chest X-ray examinations, maintained by the U.S. health care USA  National Institutes of Health. The catalog contains more than 110 thousand fluorographic images associated with 14 types of diseases, as well as 880 images in which certified radiologists identified 984 areas showing deviations.

All images were passed through the so-called convolutional neural network in order to sort all information in the images and encode data such as the type of disease, the location of the anomaly, etc.

The image is then divided into a plurality of regions to obtain local information. Thanks to this, doctors can quickly identify ailments and make diagnoses even without extensive knowledge in their field.

This algorithm surpasses modern machine learning, used to predict diseases, and, more importantly, gives an analytical picture of a computer decision in order to help radiologists better interpret the result, said Jia Li, head of research and development at cloud artificial intelligence and machine learning at Google[9]

2017

Artificial intelligence outperforms radiologists in pneumonia diagnosis

In November 2017, researchers from Stanford University presented the self-learning algorithm (the so-called neural network) CheXNet, which is able to diagnose pneumonia using lung X-rays. Scientists published the results of their work in the public domain. The received program is extremely highly specialized, but copes with its work better than professional radiologists.

CheXNet was trained on a publicly available database containing more than 100,000 chest radiographs on which 14 pathologies can be distinguished. After training, the neural network was checked: several radiologists were offered to analyze test X-rays, and the results were compared with the diagnoses of the machine. As it turned out, the computer system was able to diagnose pneumonia more accurately than a person.

Pneumonia is a dangerous and common disease, and detecting it early will help prevent many deaths; in the United States alone, pneumonia kills about 50,000 people every year. In addition, pneumonia is one of the main causes of infant mortality from infectious diseases.

The developers taught the system to mark in different colors those parts of the lungs where the machine "saw" signs of pneumonia; the brighter the color, the more likely the pathology. And after machine processing, the radiographs are viewed by the doctor, paying attention primarily to those areas that the machine marked as the hottest.

Andrew Ng, co-author of the article and former head of the research group in the field artificial intelligence at the company, Baidu believes that such systems will soon be used everywhere. Geoffrey Hinton, one of the pioneers of system development, in-depth training believes that the need to train new radiologists has disappeared, and the neural network may well cope with their functions. In addition to pneumonia, computer systems are also able to detect signs of the presence of tumors, rhythm disturbances hearts and other pathological changes on X-rays, electrocardiograms and other imaging systems.^[10]

Overview of the development of AI in radiology

In early May 2017, AuntMinnie.com published an article on the use of artificial intelligence (AI) in radiology. The publication concludes that the computer is not yet ready to replace radiologists, but is able to radically affect their activities by optimizing workflows, reducing the number of examinations and even helping to find genomic markers.

According to Dr. Eliot Siegel of the University of Maryland, over the past 30 years, thousands of works have been published on algorithms for analyzing medical images, but few of them have found application in medical practice. Those technologies that have nevertheless been adopted are rather evolutionary, and not revolutionary, he noted.

Mammographic detection systems with computer data processing have been around for more than 20 years and are essentially the only machine learning programs that are widely used in diagnostic imaging today. About 90% of mammographers use it, but with a huge degree of doubt and skepticism, and in some cases with ridicule, "Siegel said.

However, he said the use of GPUs for machine learning has led to a significant increase in computing power. This made it easier to implement the most resource-demanding computational techniques, such as deep learning, image recognition, and convolutional neural networks.

All this has great potential for working with medical images, but it requires careful labeling and detailed descriptions of images by hand, as well as a large number of imaging studies, the expert added.

Dr. Marc Kohli of the University of California, San Francisco, agrees that artificial intelligence is still far from the clinical practice of the average radiologist, despite numerous projects in Silicon Valley.

At the same time, more and more specialists in radiation diagnostics are gaining basic knowledge about machine learning and their application in practice. Academics hope the growing interest in artificial intelligence will increase investment in the field. At the same time, data processing and analysis specialists are increasingly entering many academic programs, Kolya notes.

Dr. Bradley Erickson from the Mayo Clinic in Rochester (Minnesota, USA) says that by 2017, the most common area of ​ ​ use of machine learning is speech recognition, but there are still many areas for improving the technology.

According to Dr. Luciano Prevedello of the Ohio State University Medical Center at Wexner, the development of artificial intelligence in radiology is fast, but this area is still in its infancy.

Prevedello expects AI solutions to start being used more frequently by 2019 to streamline workflows. For example, one of the Ohio clinics already uses "smart" algorithms to search for CT scans with critical readings. The use of AI for diagnosing from medical images will be possible later due to the need to conduct numerous tests before implementation, the doctor said.

Dr. Raymond Geis of the University of Colorado School of Medicine believes that AI is useful for finding patterns in data and predicting behavior based on previous models, and the first commercial AI algorithms in radiology will not analyze the images themselves. Rather, they will be deployed to produce reports based on clinicians' findings.

Commercial solutions can be used, for example, in CT scans to alert doctors to dangerous abnormalities, such as acute intracranial hemorrhage, Gase suggests.

Eliot Siegel sees great promise for artificial intelligence in specialized fields: identifying vertebral body fractures, detecting and characterizing nodes in the lungs, MRI and nuclear analysis of heart images.

Mark Kohli found it difficult to name which areas will be most important for the clinical use of artificial intelligence in radiology, but expressed the opinion that the success of projects will depend on many factors, including the correctness of integration and inspections, as well as on the decisions of regulators.

Applications that do not need regulatory approvals will evolve faster, since the authorities do not have a debugged mechanism for developers of PACS and other similar systems to offer AI software created by third-party developers due to strict requirements for code writing, documentation and testing.

Another problem, according to Siegel, is the high system requirements for computers that run machine learning algorithms. Cloud services could be a solution to this problem, but many radiologists do not trust them and do not actively use them, he said.^[11]

Bradley Erickson is confident that artificial intelligence will primarily be useful for displaying genomic or diagnostic properties that radiologists do not see today.

We and others publish studies showing that deep learning can predict genomic markers with high accuracy from conventional CT and MRI images, even if people know little or nothing about these signs, "Erickson said.[12]

Spanish AI software for disease recognition from X-rays

As part of the European Society of Radiology (ECR-2017) congress, which was held in Vienna from March 1 to 5, the Spanish Institute of Medical Research La Fe (Medical Research Institute La Fe) presented software that uses artificial intelligence for primary disease detection from X-rays. Read more here.

See also

• Telemedicine
• Telemedicine (Russian market)
• Telemedicine (Global Market)
• Remote monitoring of patient health

Notes