Professor Dr Dirk Hempel of Onkologisches Zentrum describes the Oncoguide system for diagnosis and treatment of chronic myeloid leukaemia/myeloproliferative neoplasms and myelodysplastic syndromes
As a result of diagnostic and therapeutic advances, cancer patients live significantly longer than they did 10-20 years ago. It is therefore to be assumed that more malignant diseases will in the future increasingly take a chronic course. In this case, disease monitoring, from which therapeutic modifications result, is necessary. All available therapy options must be evaluated carefully and continuously in order to ensure optimal therapy management. The rapidly growing possibilities of so-called ‘personalised medicine’ require the continuous adjustment of the current molecular-genetic tumour changes with the available therapeutic options. The rapid growth of knowledge in this field requires the development of computer-based assistance systems. A further aspect is the complex drug interactions in modern tumour therapy, which are difficult to manage without such a system.
Clinical practice guidelines (CPGs) help to translate the reviewing and evaluation of the huge number of publications into concrete diagnostic, treatment and follow-up recommendations. CPGs represent the current state of research. Beforehand, they were passive disseminations (e.g. distributed via print media), but this classical dissemination doesn’t assist the physician in the adaptation of CPG into the daily diagnostic and treatment algorithm to the given boundary conditions (patient, equipment, medical experience).
The project aims to develop and implement a computerised interactive assistance system for the diagnosis and treatment of chronic myeloid leukaemia (CML)/myeloproliferative neoplasms (MPN) and myelodysplastic syndromes (MDS). The system is based on modulated, established medical guidelines with reference to the individual patient information. The user is navigated through the complex recommendations of current guidelines and decision trees of the CML/MPN and MDS, similar to a car navigation system.
In addition to the guidelines, recommendations on current evidence-based research results will be incorporated into the expert system. The medical user is hereby offered a systematic decision support. In addition, the decisions are documented and visualised in an intuitive user interface. The expert system is also able to indicate therapy studies that are currently recruiting.
The future system is planned to be self-learning by assessing the decision criteria, which it will adjust by increasing its usage and through possible readjustments.
Technical implementation and advantages
From a technical perspective, the knowledge-based system is implemented as a client-server architecture. The server acts as a central data storage, in the form of a database, for example. As a client, internet browsers can be used so it is possible to centrally store the accumulated knowledge and to present the current research results to each user. The advantage of such a technical solution compared to guideline books is rapidly adaptable knowledge on new research results, as well as the implementation of a self-learning system. The system design has to pay particular attention to an intuitive user interface. On the server’s side, the knowledge of guidelines and interviews with experts have to be formalised in an appropriate manner. There are approaches based on an oncological or logic-based modelling.
On the client’s side the system suggests the user-appropriate decisions for the diagnosis and further treatment of diseases. The underlying methodology is based on approaches from artificial intelligence, such as the Bayesian inference or machine learning methods.
The presented computerised interactive assistance system could help to increase the accuracy of diagnosis, treatment and follow-up of CML/MPN and MDS.
Individualised treatment options
In this project we develop a software solution that supports the treating physician in diagnostic algorithms and oncological therapy decisions. By implementing all currently available data, the system is to enable an optimal, personalised treatment.
Furthermore, the system shows the physician current clinical trials adjusted to the individual situation of the patient. Oncoguide is also configured to indicate drug interactions.
Since tumour disease is often chronic, Oncoguide indicates follow-up algorithms according to the valid guidelines. The software solution developed in the project supports oncologists in clinical and practical everyday work, operating as an assistance system, which provides individual treatment options for the doctor.
Data collection and analysis
In this project, an anonymised patient database is accessed to generate a treatment proposal. Statistical analyses and data mining approaches are used to evaluate existing known and unknown correlations between molecular tumour characteristics and the clinical course of the disease.
The data mining of the pool of molecular genetic records, e.g. by using ‘next generation sequencing or ‘liquid biopsies’, may be helpful in generating potential new predictive or prospective parameters.
The computer-assisted evaluation of larger data is already successfully used in other fields of application (for example Amazon, Spotify and Facebook).
Oncoguide extends these methods to the sensitive area of medicine and to the treating physician.
Due to the fact that guidelines and treatment methods change over time, more up-to-date patient data should be considered with a higher weight in the extraction of knowledge. For the treating physician in particular, the treatment of the individual patients can be more effective. The system allows a quick overview of possible changes or the side effects of treatment. Overall, the assistance software is intended to enable the physician to treat patients more cost-effectively. In doing so, each new patient being treated, and whose data is collected during the treatment process, expands the data stock.
Professor Dr med Dirk Hempel
Center of Oncology-Steinbeis