Precision oncology with artificial intelligence - The OnkoVisionProject

The treatment of cancer patients is facing a revolution through the use of genomic cancer medicine and immunotherapy.

In the future, tumor diseases will often no longer be treated according to their location, but according to their genomic profile. Future diagnosis and therapy decisions will therefore be based essentially on the patient's molecular genetic tumor analyzes (Weissleder, Pittet 2008). With these and other approaches, cancers that were once considered incurable can be controlled for a long time with a good quality of life. Cancer would become a chronic disease. This oncological precision medicine or targeted therapy is specifically about examining the genetic makeup of the cancer individually for each patient (multi-gene examination) in order to identify the trigger of the disease, i.e. the defective driver gene, if possible. If this succeeds, targeted treatment of the tumor disease, tailored to the patient, is possible (precision medicine).

The development of this personalized medicine has already achieved such a dynamic that it will quickly find its way into routine care and must therefore be widely available. Likewise, diagnostics will quickly reach a level that can no longer be handled by a few urban centers in Bavaria.

To enable this precision medicine, enormous amounts of data have to be saved and processed. Some of this data is already being collected today, but not research and provided in a supply-compatible manner.

Data generated in rural areas in particular have so far hardly been processed in a targeted manner. The potential resulting from the basic availability of the 2D data thus remains largely unused. As a result, innovative medical concepts of precision oncology are currently not available nationwide in Germany. It is estimated that only around 30% of cancer patients currently benefit from these options. However, intelligent preparation and integration of these “data treasures” is the key to precision medicine and therefore a decisive factor in the fight against cancer. Molecular genetic analyzes are also suitable for identifying new biomarkers as well as prognostic and predictive indicators in "health services research". For this purpose, methods of artificial intelligence (AI) and in particular deep learning are necessary in order to search specifically for genetic changes and patterns and to compare them with clinical courses and radiological findings. By processing huge amounts of data and high computing power, artificial intelligence enables an exact interpretation of molecular genetic diagnostics. This paves the way for making expert knowledge and tailor-made therapies widely available in routine care.

To take these developments into account, the following main objectives were identified for the present project:

1.Design of an integrated approach to enable the comprehensive availability of molecular tumor boards for patient-centered therapy at every tumor center in Bavaria

2. Targeted use of the collected "Real World Data" for health services research - in particular for drug approval and effectiveness assessment

3. Integration of the available data in a platform for secure and data protection-compliant analysis and the exchange of health data between distributed actors