Artificial Intelligence in Healthcare — Hype or Real Breakthrough?

In previous years, medicine was mainly focused on the treatment of acute diseases, but today it pays more attention to chronic ones like obesity, depression and diabetes.

Detection of heart failure, autoimmune disorders and cancer in the early stages saves millions of lives all over the world, but on the other hand the doctor’s task becomes much more complicated.

Even the most experienced professionals cannot make the right decision in a blink of an eye, as with every hour the volume of medical data, which has to be analyzed and taken into consideration, grows rapidly.

Today, in order to quickly solve many health-related problems, doctors need to use artificial intelligence to make the right decisions.

What is Artificial Intelligence?

When saying “Artificial Intelligence” (AI) we mean the ability of the machine to simulate rational human behavior.

Artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. In computer science AI research is defined as the study of “intelligent agents”: any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, the term “artificial intelligence” is applied when a machine mimics “cognitive” functions that humans associate with other human minds, such as “learning” and “problem solving”. Source: Wikipedia.org

Today there are two main concepts of AI in healthcare: expert systems and neural networks.

The first expert systems were created a long time ago — in the 1970’s. They were introduced by the Stanford Heuristic Programming Project led by Edward Feigenbaum, who is sometimes termed the “father of expert systems”. The first expert system was made to help with diagnosing infectious diseases and identifying unknown organic molecules.

A key part of the expert system is the knowledge base — a set of information about the subject and a set of instructions applicable to facts.

Facts in the knowledge base of the expert system describe constant phenomena in a particular subject area. For example: “a healthy person has two legs.” When the system receives some information, for example: “the patient has one leg”, working memory compares this data with the one in the knowledge base. If this fact does not correspond with the one in the base, we receive a verdict: “the patient is sick.”

Building expert systems requires huge resources. To get a good expert system, you need experts in the field, knowledge engineers and programmers. The knowledge base must not only be created but it should also be constantly updated.

Today the concept of expert systems experiences a serious crisis as artificial neural networks (ANNs) are becoming more and more popular due to the possibility of their constant training and improvement.

The mechanism of ANN operation is based on the principle of biological neural networks. In computer form, ANN represent a graph with three or more layers of neurons connected in one way or another.

Mechanism of ANN operation

During training, input neurons receive some kind of data. Later it is processed by neurons on the inner layer, and the output neurons get certain new values.

If the obtained values do not suit the researchers, they change the weight of the connections of the neural network and re-teach it. The more data ANN receives, the more reliable responses to the requests it gives.

For example, we put the following symptoms into the system: headache, chills, fever. After analyzing the medical records of thousands of patients, ANN gives the following answer: “high probability of flu”.

An important note — the neural network does not know what is headache, chills or flu. It only finds a connection between the symptoms and the doctor’s reports in the sample data and then ranks these links according to their weight.

AI vs Conventional Computer Programs

Unlike the situation with common computer programs, when creating an artificial intelligence, the programmer does not need to know all the dependencies between the input data and the result. In areas where there are already mathematical models created (for example, for the statistical processing of medical records) there is no need to implement artificial intelligence.

The key task for artificial intelligence is to process all the provided data and find the formulas and dependencies that are not yet determined by scientists (and in some cases, they never will be).

The Possibilities of Artificial Intelligence in Healthcare

Each medical picture or anamnesis contains information that allows specialists to accurately diagnose the disease and prescribe the necessary treatment. Unfortunately, even experienced doctors can’t always see the full picture of the disease, because the data in the medical record is not structured, and the history of the disease can be too voluminous. The effectiveness of their work is also affected by fatigue and in some cases — lack of knowledge.

Skychain’s neural network localizes lung diseases using heat maps

For example, such disease as cancer can easily be cured if it is diagnosed at an early stage, but doctors can simply miss it on the x-rays (according to the statistics the doctors miss the lung cancer at its first stage in 70% of cases).

Artificial intelligence can solve this problem. AI’s for patient assessment and pre-diagnosis are already being tested in a number of US hospitals.

In many countries, you can sign up for a doctor via the Internet. However, there are many other patients, who use this feature, so sometimes you have to wait for several days, if not weeks.

Modern technologies can help us. One way to solve this problem is mobile telemedicine with AI features. Such solutions involve the use of smartphones and wearable devices (such as smart watches or fitness bracelets) to assess human health.

The data, obtained from these devices can be “fed” to special AI program, which is trained to recognize malignant tumors, diagnose tuberculosis, as well as visual impairment and “failures” in the work of other organs, including the brain.

Systems based on artificial intelligence are able to process thousands of pages of text per second when searching for the necessary information. This is extremely useful as no doctor can cope with the operational analysis of such a wide number of medical publications.

According to the Delve Health Company, every 20 minutes a new medical article appears. For example, MEDLINE catalogue added 870,000 links to different healthcare-related articles in the year 2017.

Interface of Dxplain decision support system

Therefore, doctors need decision support systems (DSS) based on artificial intelligence. DSS combines information from the patient’s medical history, health indicators, data from medical directories and the latest research. Then AI looks for dependencies, taking into account even such factors as temperature dynamics, noise level and air quality in the place of residence to provide useful information, which can be later used by doctors to prescribe treatment.

According to Judy Sewards, a top Executive of Pfizer, the process of bringing to market a new drug takes more than 12 years. Drugs are the most complex organic compounds, the search for the structure of which is carried out almost blindly.

The scientist changes the molecules in the original compound and checks the reaction of the experimental animal to the drug and so on until the optimal result is achieved. Only then clinical trials begin.

Before entering the market, the drug undergoes a lot of inspections by regulatory authorities. But even the most thorough tests do not guarantee that it will be effective. Almost half of the new anti-cancer drugs do not have a noticeable therapeutic effect.

AtomNet neural network predicting candidate treatments for Ebola infection.

Artificial intelligence will be used to create new drugs. In the future, scientists will be able to set the desired properties of the chemical compound, and the computer will form the necessary molecular structure.

Market Potential of Artificial Intelligence in Healthcare

According to Frost & Sullivan Company, the revenues of companies, which operate in the healthcare AI market will reach 6.1 billion dollars by 2021.

The Research and Markets Company predicts more modest future for AI in healthcare: by 2020 the market will grow to 5.05 billion.

According to R&M agency, the fastest growing segment of AI implementation will be healthcare.

Optimists from Frost & Sullivan predict that by 2025, the AI system will be used in 90% of clinics in the United States and about 60% of the largest hospitals in the world.

Join Skychain on social media: Twitter, Facebook, Telegram

Egor Chertov, Skychain team

Blockchain infrastructure aimed to host, train and use artificial intelligence (AI) in healthcare. Our website: https://skychain.global/