Assistance Programmes in Medicine
The Digital Expert
Specialist computer programmes are being used increasingly to help doctors diagnose diseases and recommend treatments.
Making sure your know-how is always up to date isn’t easy – not even for doctors. Because nowadays the stock of medical knowledge doubles every 75 days. [1] That’s a huge flood of information for anyone to keep on top of. Computers can provide help here. Digital expert systems guide physicians through the flood of recommended therapies and search for the appropriate treatment approach. Or, functioning as an “automatic watchdog”, they check patients’ laboratory data and sound the alarm if any problems flash up. We’ll give you two examples of how such assistance programmes are developed and where they can be used.
Example 1: Using algorithms against cancer
Delivering the same therapy to every patient with the same disease is a thing of the past. Today, doctors tailor their treatments to their patients’ needs. In the case of diseases such as blood, breast or colon cancer, different patients might receive different, individualised medication. In order to find a suitable therapy, doctors might, for instance, use “molecular profiling” to examine a so-called solid tumour, which can originate in different organs. More than 33,000 gene variants can be displayed and tested for mutations caused by the cancer – and this number is growing rapidly [https://www.foundationmedicine...].
Some of the gene mutations are relevant to the course of a disease, while others are not. “This makes it virtually impossible for us to know the right therapy straight off from the word go,” says Professor Dirk Hempel, head of the Cancer Center Donauwörth in southern Germany. At a number of linked treatment centres in Bavaria, he is currently testing a digital expert system named “HämaNavigation”. The latter uses a database to check which therapies are suitable for which gene mutations. Based on guiding principles and the latest research findings, the programme suggests a treatment and refers doctors to the relevant study papers. The guiding principles that point the doctors in the right direction are based on scientific knowledge and tried and tested procedures. They have been developed by experts and medical societies.
You can learn how the guiding principles for expert systems are developed and then carried over into the systems themselves by reading these two studies (in English): https://www.esmo.org/Guideline... and https://www.nccn.org/professio....
Computer-aided prognoses
The digital assistant was jointly developed at the Steinbeis Transfer Institute for Hematology-Oncology and the Fraunhofer Institute for Optronics, Systems Technologies and Image Exploitation IOSB. The computer programme is being tested on patients as part of routine care, for example on those suffering from so-called myelodysplastic syndrome, in which the bone marrow has been damaged and the process of blood formation no longer functions properly. “In some patients, this leads relatively quickly to acute leukaemia; in others the syndrome tends to be benign,” says Hempel. “The more gene mutations, the higher the probability that the disease will take a critical course.” The system calculates how high the risk is for each patient. If necessary, doctors can administer a drug at an early stage to prevent the onset of leukaemia.
After initial testing, Hempel draws a positive conclusion: “Medicine is currently being revolutionised by immune and gene therapies. Digital systems like HämaNavigation are essential to keep up with this development.” HämaNavigation has been in use since January 2019. Whether the system actually benefits the patients treated has been under evaluation since November 2019.
Example 2: Faster therapies thanks to artificial intelligence
In Germany alone, around 260,000 people suffer a stroke every year. [4] The earlier those affected receive the right medical care, the greater their chances of avoiding brain damage. Every minute counts when it comes to preventing tissue death. For this reason, the Dutch start-up Nico.lab, for example, is developing a digital assistant to help doctors make the right diagnosis and initiate the appropriate treatment as fast as possible.
The so-called “StrokeViewer” is designed to support physicians in evaluating more quickly the brain images of affected patients. In order to determine the cause of the stroke, images are usually taken in a computer tomograph. However, early signs of a stroke are not always easy to detect in a brain scan – even for experienced doctors. A dark grey spot on the brain image indicates a stroke caused by a blood clot. A white spot indicates a brain haemorrhage. Choosing the wrong therapy can endanger a patient’s life. [5]
The Nico.lab software evaluates the images from the computer tomograph with the help of “artificial intelligence”. This method is faster than was previously possible with manual evaluation by doctors. To enable this, algorithms were trained with data from stroke patients. The system is now able to detect changes in the brain scan and to indicate, for example, where a blood clot has formed.
The software also calculates the amount of blood flowing through the major brain arteries. This enables doctors to assess whether any other blood vessels are there to ensure that the affected brain areas are still being supplied with blood despite the clot. The better these so-called collateral values, the more likely it is that patients will benefit from a certain treatment. In order for doctors to be able to estimate the values, patients have until now had to be administered a contrast agent. [6] The “StrokeViewer” from Nico.Lab is currently still being further developed and tested in clinical studies. Nevertheless, 15 hospitals in Amsterdam have already expressed interest. [7]
Sources and bibliography
[1] https://healthcare-in-europe.c...
[2] Interactive Decision Support: A Framework to Improve Diagnostic Processes of Cancerous Diseases Using Bayesian Networks (Patrick Philipp, Sebastian Robert, Dirk Hempel, Jürgen Beyerer), 2018
[3] Modeling of Clinical Practice Guidelines for an Interactive Decision Support Using Ontologies (Patrick Philipp, Marie Bommersheim, Sebastian Robert, Hempel Dirk, Jürgen Beyerer), 2018
[4] https://www.dsg-info.de/presse...
[5] https://www.spektrum.de/magazi...