Demographic change and medical advancements are moving forward at breakneck speeds. Not only is the number of people suffering from common diseases such as diabetes, asthma, heart failure and depression on the rise, the number of cancer sufferers is also increasing. Thanks to targeted areas of research in oncology, the treatment options are growing as well – but this means that the field is becoming more complex and it is increasingly difficult to keep an overview. Doctors are now wondering which treatments are best for patients with certain physical constitutions or genetic predispositions. The Steinbeis Transfer Institute of Clinical Hematology-Oncology at Steinbeis University Berlin is working with partners to develop software showing guideline overviews that will help to make doctors’ lives easier.
The past two decades have seen hundreds of thousands of publications come out, highlighting the various illness indications found in a vast array of medical areas. Add to that the results of experimental studies, which are published on an almost daily basis. Doctors have to take both areas into consideration in their decision-making processes. What’s more, the number of diagnostic tests for assessing predictive therapy factors is also growing rapidly – making it harder to keep up to speed with advancements using conventional media. How should or can a doctor make decisions based on the latest medical findings?
Developing guidelines and directing therapies in line with these condensed recommendations is the method of choice. Evidence-based guidelines are derived from systematically compiled and processed information in the literature. They are regularly updated or contain notes on their period of validity. After evaluating lab values and patient parameters, doctors refer to the guidelines to define the first stages of treatment. Depending on how treatment goes, further possible treatment is decided upon. Until now, doctors have had to do this by referring to printed guidelines, or have had to look them up online. Having software with the guidelines would make the work of a doctor much easier. In addition, decisions could be better documented and more easily traced.
The Steinbeis Transfer Institute of Clinical Hematology-Oncology and the Fraunhofer Institute of Optronics, Systems Technology and Image Processing (Karlsruhe) hope to develop an oncological expert system in a homogenous and transparent outpatient medical setting, and then test it in daily practice with the aim of implementing the solution in the broader field of oncology care. The expert system will initially be developed and tested for a limited and manageable range of indications. The main field work will be carried out at regional, well-networked outpatient oncology centers north of Munich (Donauwörth-Dachau and Freising-Fürstenfeldbruck). The oncology centers treat nearly 3,500 tumor patients each quarter in close cooperation with nearby hospitals. This makes these structures ideal for delivering important practical and theoretical input on the development of the system. The medical database needed for the expert system can also be used to measure the standard of treatment offered in outpatient oncology. This is very important for specialized outpatient care.
The knowledge-based system will be implemented in a client-server architecture, where the server acts as a central data source, and something like an Internet browser can be used as a client. A technical solution of this nature would make it easier to update content as new research results emerge, giving it a clear advantage over standard guideline publications. It will allow doctors with the right qualifications to update content (for example findings presented at a tumor conference) which will then be available to other doctors in general practices. On the server side, the knowledge from guidelines and interviews with experts needs to formalized in some way. This means implementing approaches based on ontological or logic-based modeling. Quality management would make it possible to separate “confirmed” knowledge from new knowledge that has yet to be confirmed. On the client side, the system can recommend tests and support doctors making a decision on an individual patient by showing their previous entries made to the system. The underlying system is based on artificial intelligence techniques, such as reasoning, Bayesian inference, and machine learning.
The project partners are planning a multiphase project. Since developing a self-learning expert system for oncology is such a complex technical, medical, and even politically sensitive project, the project partners would like to start by developing software with guidelines on indications. In a further step, they envision extending the program to access the data in other databases – for example from tumor centers in the region or cancer registers.
The Steinbeis Transfer Institute of Clinical Hematology-Oncology will primarily work on the treatment research in outpatient oncology, with a particular focus on outpatient specialist care. It will also manage the transfer of results into practice and provide support research into ethical issues. In addition, a scientific advisory committee will work on the project with experts from the field of oncology as well as political and business partners.