Artificial intelligence in medicine
Big Data as the basis for reliable diagnoses, machine learning to select the best treatment, AI systems for accelerated drug development – The importance of artificial intelligence is also growing in the field of medicine. What is the potential of self-learning algorithms in the healthcare sector?
Fast and accurate: medical data analysis with AI
“>Artificial intelligence is prevalent in many areas today. In the field of medicine, powerful AI software has become indispensable. AI is of particular help in the following areas:
- Diagnosis (determination of diseases)
- Therapy (selection of the treatment method)
- Development of drugs
- Editing genes (gene therapy)
Big Data helps provide fast and reliable diagnoses. This is of particular benefit to patients with rare diseases, because it helps prevent long periods of suffering. Indeed, the rarer a disease, the less likely it is that patients will find a doctor with specialized knowledge. Discovering the true cause can therefore take a long time. A patient suffering from a rare disease sees an average of seven doctors and has to wait an average of five years for the diagnosis to be confirmed. To identify the complex clinical patterns of such diseases, AI helps those physicians who are not experts for the disease in question.
Artificial intelligence needs training. High quality training data is provided by the Crowd. Ask clickworker.
AI prevails over doctors
The benefits of using artificial intelligence in diagnostics are obvious. Computers can analyze data much faster and more accurately. And the algorithms will become even better, the greater the amount of training data that is made available to them. In some sub-areas, AI already achieves better results than human analysis.
- In a test conducted by the German Cancer Research Center, AI more often detected suspicious skin changes than real doctors.
- AI also helps in the early detection of colon cancer. Harmless growth or dangerous precancerous stage? Computer-aided systems check changes at lightning speed using hundreds of individual characteristics and thus avoid unnecessary interventions.
However, one thing is clear: AI systems cannot substitute medical decisions, they can only support them. Artificial intelligence provides physicians with highly specialized tools. But none of these tools can take the place of a doctor. The evaluation of extensive amounts of data enables the physician to select the appropriate diagnosis or therapy from the suggestions made by an AI system. This saves time and also ensures that all possible alternatives (e.g. for therapies) are carefully assessed by the expert.
More effective drugs thanks to AI?
Pharmaceutical research essentially consists of revealing correlations between actual effectiveness and disease treatment. When analyzing these correlations, artificial intelligence can perform many routine tasks — usually better and more accurately than humans. A decisive advantage here is the acceleration of active ingredient production. While this takes four years on average, an initial success with artificial intelligence was achieved within only twelve months.
This first active ingredient developed using AI is called DSP-1181. It is based on the machine comparison and analysis of different molecular compounds. This was achieved by using artificial intelligence to link the extensive information in databases with the patients’ unique genetic factors.
Dreams of the future: Efficient AI in medicine
Artificial intelligence evaluates certain signals and provides information about the presence of diseases. The possibilities for the application of AI are basically unlimited. Today, AI is being used in many medical fields – usually to complement decisions or as a basis for medical treatment.
- Early detection of epidemics by means of an Internet search for the increased occurrence of certain keywords,
- Suspicion of heart or brain disease by evaluating ECG or EEG signals,
- Recognizing symptoms of depression through movement patterns and speech.
This is where image, speech and text recognition play a key role. All the applications existing today have one thing in common, they are based on so-called weak artificial intelligence. These include systems that are capable of learning and thus of self-optimization, but ultimately have a limited range of application. They are always programmed to solve comparable tasks.
The so-called strong AI, on the other hand, not only functions within the specifications, but is creative in a certain sense — and thus “intelligent” in the human sense. It automatically recognizes problems (for example, new diseases) and develops solutions. This could provide new diagnostic and therapeutic techniques for medicine. However, most scientists believe that strong AI is still a long way off.
The acceptance of AI by the general public
It goes without saying that the use of artificial intelligence in healthcare can only work if patients also approve of it. Is there a general acceptance of AI systems in the medical field? According to a 2019 survey by bitcom.org many respondents are open-minded about the use of AI in medicine:
- Just under a third of those surveyed (31%) would regularly consult AI systems to obtain a second opinion.
- Nearly two thirds (61%) believe that using AI for routine procedures will give physicians more time for their patients.
- Some 39 percent of those surveyed say they are in favor of physicians having their diagnoses checked by AI on a general basis.
These results show a general trend: Acceptance is high as long as AI helps economize resources. However, although it is unrealistic, many people are afraid of being remotely controlled by AI. The more convincing the results of machine learning become for medicine, the sooner unfounded fears will disappear.
Can AI replace the doctor?
The success of AI-based data analysis is astounding in the field of medicine. And the more data is available, the closer the possibility of identifying specific risks in terms of individual disease prophylaxis at an early stage, thus preventing health problems in advance. This requires a lot of data — and people must be willing to disclose this data. There is a basic acceptance for medical AI applications, but it will be years before an AI robot replaces the human doctor.