According to Venture Scanner, the amount of companies engaged in AI solutions for healthcare has exceeded 2000 worldwide as of the beginning of 2019. One of them is Philips. Its CEO, Frans van Houten, asserts thatdisease prevention is the future of medicine. Therefore, artificial intelligence should diagnose diseases as accurately as possible in this sector. The article reveals what services already operate and how successful they are.
IBM Corporation has developed a platform called Watson Health aimed at the improvement of healthcare services. A range of its solutions includes several services, in particular a diagnostic one: Watson for Oncology, a cognitive program assisting in treatment planning for various types of cancer. It analyzes large volumes of medical literature, compares different factors, and draws analogies. Based on these data and after examining patient's history, the system establishes potential diagnoses and treatment options, but the ultimate decision is always announced by the doctor.
In 2016, Watson for Oncology detected wrong diagnosis and give the correct one: in very deed, the sixty-year-old patient had a rare form of leukemia. Hereby, the program saved life of this woman. Currently, Watson for Oncology is applied at hospitals in the USA, Thailand, and India.
IBM strives for expanding Watson Health possibilities so that its databases won’t be limited to oncology. Thus, in 2015, the corporation acquired Merge Healthcare, a company developing solutions for collection, storage, and examination of medical images. Its systems are installed in 7500 US clinics as well as in world’s healthcare companies and pharmaceutical firms. IBM spent 1 billion dollars to purchase the firm and depersonalize all patients’ information. Watson Health analyzes these data using a deep learning method and finds the needle in a haystack, as Deborah DiSanzo, CEO at IBM Watson Health, stresses.
Google has designed its own version of AI for medical examination and improvement of nursing technologies. DeepMind Health is similar to IBM’s supercomputer and is applied at several UK hospitals. For example, this technology allows Moorfields Eye Hospital to decipher 3000 OCT eye images every week. DeepMind Health, within a short time, reads out information, can diagnose macular degeneration and diabetic retinitis, as well as provides doctors with recommendations on ophthalmic disease treatment.
Besides, Google’s subsidiary focusing on DeepMind Health solutions is investigating 500 case studies of head and neck cancers in collaboration with University College London Hospitals.
Earlier, DeepMind Health was a part of the UK research team – DeepMind, but it was acquired by Google in 2014. This fact has led to questions about data integrity, but the corporation ensures that they are enciphered, protected, and processed in accordance with the law.
In 2017, a Singapore-based technological company, Hanalytics, developed BioMind AI technology along with Beijing-based Tiantan Hospital. During the months, BioMind was examining dozens of thousands of medical images collected over decades. To show its capacity, developers held a competition. A single neural network and 15 specialists were competing for the speed and accuracy of encephaloma diagnosing based on data of 225 medical records. Artificial intelligence completed the task in 15 minutes, giving the accuracy of 83%. Doctors got through all the information in half an hour, providing correct conclusions only in 63% of cases.
AI healthcare technologies are a key part of the government strategic plan ‘Made in China 2025’. According to OECD Data, less than two doctors (1.8) accounted for one thousand people in China as of 2015. Artificial intelligence will help the Chinese heavy medical system to get rid of routine as well as to increase the efficiency of treatment and diagnostics accuracy. The Government of China encourages the population to carry health monitoring devices so that technologies will have more opportunities to obtain information for education and analysis.
BioMind has been already used at Tiantan Hospital since December 2018. Its specialists believe that the machine does not replace doctors but speeds up their work and tirelessly fills out blind spots in research. This view is shared by Raymond Moh, CEO at Hanalytics.
Zebra Medical Vision
Zebra Medical Vision has produced Zebra AI1, a device that combines pulmonology analytics and visualization. It is based on the deep neural network library, cuDNN, containing hundreds of thousands of CT, MRT, U/S, and other medical images. The platform reads them and can diagnose hepatic, pulmonary, vascular heart, and bone diseases, as well as assess patients’ risk profiles. Afterwards, Zebra AI1 makes a conclusion and gives it to doctors for the ultimate decision.
The company’s blog states that Zebra services are used by 50 hospitals across the globe and one can order a research on their website for 1 dollar.
One more platform for medical visualization by the same-name startup. The Arterys system combines artificial intelligence and data cloud storage technologies, and diagnostic imaging. Put simply, it allows to conduct MRT and visualize the blood flow system into multi-dimensional models. By the way, cardiac scanning, for instance, takes only 10 minutes unlike standard 30–60 minutes. Arterys analyzes images and can diagnose not only heart diseases, but also pulmonary, hepatic, and breast diseases.
Arterys solves the privacy issue using the PHI service that “cuts out” all personal information at the preparation stage, i.e. when a hospital sends a request for research. The company already operates in the USA, Canada, and France.
Learning mHealth services usually cope with medical conclusions just as well or better as doctors similarly to Watson for Oncology and BioMind. Arterys’s CEO believes that doctors are physically unable to process all data on their own. Thus, they can rely on artificial intelligence and neural networks. Both IT specialists and health professionals see eye to eye on this matter.
Nevertheless, it is impossible to establish a diagnosis without human involvement now. Dr. Wang Chongqing from China says that people and machines are good in different things. Determining a diagnosis, doctors take into account the whole range of factors, including patient’s gender, age, living conditions, ecological features, etc.
Despite the fact that AI systems increase the diagnostics accuracy by 30–40%, the final decision is always accepted by a human being. Dmitry Dozhdev, the head of CoBrain, says that AI won’t replace doctors in diagnostics issues, so doctors still bear the responsibility for decisions. The head of Fusion Fund confirms these words, adding that neural networks just speed up hospital operations and prevent specialists from forgetting something important.
Robert Merkel, Vice President at IBM Watson, reports that one electronic medical record contains, on the average, 400 GB of data. If one adds person’s genetic information, this volume will grow up to 6 TB. Well, AI is essential in cross comparison of patient’s personal data and medical knowledge.
Experts will discuss artificial intelligence technologies at Moscow’s AI Conference.