In previous articles, we have already talked about the prospects for artificial intelligence. It's high time to talk about its implementation in individual industries and we will start with perhaps the most socially significant - healthcare. Medicine and healthcare are already considered one of the strategic and promising areas in terms of the effective implementation of AI. The use of AI can massively increase the accuracy of diagnosis, make life easier for patients with various diseases, increase the speed of development as well as release of new drugs, and much more.
For the medtech business, this area is also a very tasty piece of cake. According to international studies, the use of artificial intelligence in healthcare can significantly increase the gross margin of companies in the healthcare industry. According to Global Market Insights, by 2025 the total AI market in healthcare will reach $28 billion with a CAGR of more than 45.1%, and the AI market for healthcare imaging and diagnostics will reach $2.5 billion. Events like a coronavirus pandemic that took the world by surprise at the beginning of 2020, only confirmed the need and relevance of work in this direction. Let's find out what developments in the field of AI for healthcare already exist, which are planned and what are the prospects for the advancement of this industry.
How is AI in healthcare developing?
The main motivation for the development of AI in healthcare is the increase in costs and the corresponding need to limit them. Also, there is a problem of the diagnosis quality: about 20-30% of medical research is inaccurate. Don’t forget about the desire to standardize and automate routine functions up to the creation of self-guided diagnostic models.
Another obvious incentive for the development of AI is the huge amount of data that is generated by all kinds of healthcare devices and information systems. At the same time, after 3 months of storage, less than 15% of the healthcare data is demanded by doctors. AI seems to be great solution because of the ability to give meaning and added value to the accumulated information. However, let's move on from lengthy reflections to the main directions and special cases of this industry.
AI algorithms are already in full use in medical practice, helping doctors identify diseases and prescribe treatment. It can be difficult for a doctor to correctly diagnose a disease, especially if he does not have too much practice or a specific case is far from his professional experience. Here, artificial intelligence can come to the rescue, having access to databases with thousands and millions of case histories (and other ordered information). With the help of machine learning algorithms, it classifies a specific case, quickly scans the scientific literature that has been published over a certain time interval on a specific topic, studies similar cases that are available, and offers a treatment plan.
An example is IBM Watson for Oncology, the IBM Watson supercomputer utilization program for determining the optimal evidence-based cancer treatment strategy.
The use of AI in the treatment of cancer
China, in its turn, developed a system based on the Tianhe-1 supercomputer for identifying patients with COVID-19 using computed tomography based on AI technology. The use of computer algorithms is necessary in order to distinguish the new coronavirus from other varieties of pneumonia. The system was developed by the China Cancer Association in conjunction with the National Supercomputer Center in a central subordinate city of Tianjin.
In addition, in the United States, the probability of a medical error, according to some estimates, is 9.5%. Automation will give the doctors extra time, which they can spend on studying the patient’s illness and establishing the most accurate diagnosis. According to pathology specialist Andy Beck (Harvard Medical School), the use of AI technology will reduce the level of error in diagnosis by 85%.
Artificial intelligence has proven particularly good in predicting diseases. Thus, a system developed at the University of Louisiana is capable of predicting epileptic seizures with an accuracy of 99.6% an hour before the onset of the main symptoms. A similar solution by IBM and Astra Zeneca warns the patients about the risk of heart attack and stroke by analyzing the characteristics of their body. Geisinger electrocardiogram neural network can predict heart rhythm failures more accurately by professional cardiologists.
And recently, a team of scientists taught the Deep Gestalt algorithm to determine genetic diseases from human photography. The experiments showed that in 90% of cases, one of the ten most appropriate diagnoses of AI was “correct”. One of the first AI mass deployment projects was implemented by Ping An Healthcare and Technology, the Chinese online healthcare provider. At the end of 2018, the company announced the installation of several thousand AI clinics in China. These are something like telephone booths where you can consult with a virtual doctor. Then the received information is checked by an "alive" doctor, who makes a final diagnosis and writes out a prescription.
Operations and Assistance
Developers use AI technology to create a wide range of smart assistants: from personal doctors to robotic surgeons. For example, the American robot-assisted surgical system da Vinci operates in more than one hundred clinics around the world. One of the “arms” of the robot holds a video camera transmitting the image of the area being operated, the other two reproduce the movements made by the surgeon in real time, and the fourth “arm” acts as an assistant to the surgeon. A similar Senhance system, developed in the United States, is no less successful in minimally invasive surgery.
There are also examples of virtual assistants aimed at treating mental health. Woebot is a chatbot for dealing with depressive thoughts and conditions from psychologists at Stanford University in collaboration with AI experts. It works on the basis of cognitive-behavioral therapy, which can change behavioral patterns and destructive stereotypes. A similar application is also developed by the Indian startup Wysa.
The experience of pharmaceutical companies shows that it takes approximately 12 years from the beginning of preclinical trials to the approval of the drug and the treatment of patients. At the same time, only 0.1% of “candidate drugs” come to clinical tests. Approval is received by 20% of them.
AI systems can help resolve this situation and accelerate the release of new drugs. Pharmaceutical giants like Sanofi or Novartis are already resorting to the help of startups developing healthcare innovations in order to seek new medicines. Biochemicals manufacturer Roche has acquired Flatiron Health, a company that uses machine learning to process data.
Since 2012, Atomwise startup has been using neural networks to search for more effective drug formulas. His AtomNet deep learning system checks 10 million chemical compounds daily, predicting which ones will interact best. A similar algorithm is used by the biopharmaceutical company Berg Health.
Processing huge amounts of healthcare information
Ideally, the doctor must constantly increase the level of subject knowledge, keep abreast of modern treatment practices - however, it is almost impossible to study the entire body of publications that are regularly published in subject journals - even if we’re talking about specialists.
AI technologies in combination with search engines are able to help in this situation. Such a solution was developed by scientists from the American research center RAND, engaged in the analysis of strategic problems. They taught the system to search in huge volumes of information for keywords and terms related to the subject of the request.
During tests, this topic was data on gout, low bone density, and osteoarthritis of the knee. The algorithm was able to reduce the number of relevant articles of interest to doctors by 67–83%. According to the developers, the system skipped only two articles that would be selected by people, but none of them contained critical information. The accuracy of the machine learning algorithm was 96%.
The readiness of medicine to change the usual way of life, first of all, depends on the willingness to apply solutions from other industries, including IT. Although creating your own AI solution is a rather difficult task, it is still quite possible, whether you are a clinic or a healthcare startup. The main thing in this situation is to carefully select the specialists who will implement this solution. Do not forget - we are talking about human lives!