Can machine learning learn to make diagnoses

Algorithms and AI in Medicine: Examples from Practice

Recognizing diseases, making diagnoses: software programs, machine learning and other technologies mean immense progress for medicine. That is why they are already being used in many areas to improve health care.

Everyday software

Algorithms are part of every software or app. As a result, they have long since become part of the everyday life of many consumers. This also applies to many devices that are intended to maintain health and well-being: Internet-enabled scales and blood pressure monitors, smartwatches with integrated heart rate monitors, fitness bracelets and other so-called wearables for recording various activities. There are also apps that count calories, monitor your own insulin levels or the fertility cycle.

What these product examples and applications have in common is that their functionality is relatively simple: They collect data and process it in such a way that it presents useful information for their users. Mostly it is about a certain category of data on a clearly defined aspect of health.

AI programs in medical use

In addition, applications for more comprehensive algorithms are being worked on in professional health care. Machine learning is also used in some cases. These include software programs for the prevention and early detection of diseases, to support decisions when diagnoses are made, or to select and carry out therapies.

An example of the use of AI in medicine are the so-called imaging proceduressuch as x-rays, computed tomography (CT) or retinal scans. With the help of machine learning processes, programs are “trained” to recognize indications of diseases in patient images. In this way, certain heart diseases can be identified using images of the retina, for example. Other algorithms, for example, display images of the lungs directly when they are output in such a way that the radiologists responsible are more likely to notice certain findings. For this purpose, the algorithms are “taught” what a healthy lung looks like and how known diseases show up. When presenting the images, they then particularly emphasize those elements that indicate a disease.

A video from SWR Odysso explains why medicine uses imaging processes in brain research - and what limits these processes have.

Algorithms can also recognize patterns in other data: in Copenhagen, for example, incoming calls to the emergency call center are analyzed by AI software. If the results of the Voice analysis indicate a heart attack, this is signaled to the people who take the emergency calls.

Chatbots as a diagnostic aid

For some time now, so-called Chatbots available. They offer both doctors and patients the opportunity to communicate with an artificial intelligence by entering text. Such AI chatbots are usually downloaded in the form of an app and are there to assess the symptoms entered by the users. Depending on the combination of health complaints, such apps display possible diagnoses and show how likely the system considers each individual diagnosis option.

The AI ​​consists of algorithms that can process written language and look up appropriate information in a database on the basis of the entries made. The result that the AI ​​comes to depends on various factors and can be further developed through machine learning.

Software and chatbots are not only changing the way doctors work every day. New technologies also have an impact on the role of patients - and that is what the next article is about.