
It is a problem the medical community knows well: not every hospital and not every doctor’s office are equipped with the same level of know-how. Some hospitals have specialised in certain diseases and can offer better treatment options than others. The team under the leadership of principal investigator Dieter Maier wants to remedy this situation and make medical expertise more widely available. In the context of the international research project SysMed COPD, which is co-funded by the Austrian Science Fund FWF, the researchers are developing computerised methods to make the specialised knowledge of a lung rehabilitation clinic available to other medical institutions with the help of machine learning. The project’s focus is on the treatment of smoker’s lung. In spite of many measures to discourage or ban smoking, the problem remains acute. Worldwide, more than five trillion cigarettes are smoked every year, and seven million people die of the consequences every year.
“From a medical point of view, smoker’s lung is a disease caused by smokers inhaling toxins and tiny particles,” Dieter Maier explains. “Smoking is the main cause of the disease in our part of the world, but there can also be other causes.” It is a complex undertaking to define and delineate the clinical picture of the disease, which medical language calls “chronic obstructive pulmonary disease” or COPD for short. “There are two sides to the problem. On the one hand, this disease affects only 20 per cent of smokers. We would like to be able to predict who will be affected,” Maier says. “The second issue is the range of different therapeutic approaches to COPD. At the moment it is very difficult to predict which of these options will have a positive effect on an individual patient. That is what we are working on.”
The difficulty lies in the different ways in which COPD becomes manifest, notes Maier: “There are people who are restricted in their breathing and can only take in a certain amount of air because the lungs become more rigid, but whose organism displays no noticeable irregularities otherwise. Others also develop heart problems. In the case of the latter, not only the mobility of the lungs changes, but also the ratio of blood flow to oxygen uptake and, thus, their cardiac function.” In addition, there are also those who develop severe psychological symptoms such as depression. “There are many different treatment options here. At the moment, decisions are a matter of experience and left to the doctors’ discretion. In our project, we are trying to capture all of that expertise through machine learning,” Maier explains.
Data from a specialist clinic
Viscovery, a subsidiary of Biomax, who are Maier’s employer, can draw on data from the CIRO clinic in the Netherlands to address this issue. Viscovery have been working with the CIRO clinic, who specialises in pulmonary rehabilitation, since 2012. “Over the years, CIRO have gathered a great deal of experience about which of the COPD rehabilitation measures are promising,” says Maier. In order to capture this knowledge, the researchers examined the course of the disease of patients at the clinic. “We analysed this data using machine learning to find out if we could identify groups of patients who might benefit particularly from specific treatment methods. Our approach was successful and has also been published,” Maier reports. Now the objective is to apply this approach to larger groups of COPD patients and to match it with the success of treatment over several years. To this end, Maier is using the data from a large clinical study by the name of COSYCONET that is coordinated by the University of Marburg.
More than 2700 people with COPD are participating in this study. Initial results are encouraging, says Maier: “In our analysis we were able to identify the groups that we had already identified with CIRO.” In their next step the researchers will investigate to what extent these groups differ in terms of long-term treatment success. One objective of the research is to implement the information gathered in a software tool that can be used in the IT infrastructure of doctors’ offices and hospitals. Such a software needs to be approved as a medical device, a category subject to particularly strict criteria.
The goal is personalised medicine
In addition to the software development, which is to be transferred into a commercial medical product after completion of this basic project, the researchers are pursuing yet another goal: they want to do better than merely mapping previously existing medical experience. “With the help of machine learning, it could be possible to make even better recommendations than the experts in their field are currently able to make worldwide,” Maier notes in summarising the hopes of the research team.
Maier sees COPD as just one of many possible areas of application for software of that kind. He refers to this as a “systems medicine” approach. In the case of cancer, in particular, it is now known that it varies from individual to individual. “Doctors are starting to sequence the cancer tissue for individual patients in order to understand what mutations exist and to tailor the treatment accordingly,” Maier reports. Over the last two or three years, this approach has also increasingly been used in clinical practice. Again, software tools are needed to coordinate individual treatment. University hospitals have the necessary IT infrastructure to implement such a solution, but other medical institutions need software support. Maier expects such methods to be widely used in the next few years.
The researcher emphasises that this is not about replacing people with machines: “Human attention is very important in medicine. What we want to achieve is to assist doctors with scientifically sound recommendations.” And this involves treatment as well as diagnosis.
International project
While it is rare for companies to receive funding for basic research projects, it is important, as Maier underlines. “We are a small company, and while we have the necessary technology, we don’t have the capacity to fund such a validation with patient data.” The basic research project, which started in 2017 and will run until June 2022, is part of an international research consortium involving the University Hospital in Marburg, Germany, as well as other German hospitals, the CIRO Clinic in the Netherlands and the Trondheim University of Technology and Natural Sciences in Norway. According to Maier, the problem of smoker’s lung will remain urgent: “In western industrialised countries the number of sufferers is falling, but this is more than offset by the international increase.”
Personal details
A molecular biologist by training, Dieter Maier has been working as a computational biologist for over 20 years. He heads the project management department at Biomax in Germany with a focus on systems medicine and scientific projects. Areas of special interest include knowledge management, modelling of signalling pathways, and data analysis.
Publications
Franssen FM, Alter P, Bar N, Benedikter BJ et al.: Personalized medicine for patients with COPD: where are we?, in: International Journal of Chronic Obstructive Pulmonary Disease, 2019
Cano I, Tényi Á, Schueller C, Wolff M et al.: The COPD Knowledge Base: enabling data analysis and computational simulation in translational COPD research, in: Journal of Translational Medicine, 2014
Vanfleteren LEGW, Spruit MA, Groenen M, Gaffron S et al.: Clusters of Comorbidities Based on Validated Objective Measurements and Systemic Inflammation in Patients with Chronic Obstructive Pulmonary disease, in: American Journal of Respiratory Critical Care Medicine, 2013
Comments (0)
Currently no comments for this article.