Automated image analysis improves cancer diagnoses in pediatric radiology

The long-term survival rate is 70 to 90 percent, but the chances of recovery in the event of recurrence are low. Diagnosis is traditionally invasive, for example through biopsies or bone marrow punctures. Radiological diagnostics using CT and MRI play an important role, but standardized, automated evaluation is still lacking.

The RACOON-RESCUE project, led by Prof. Diane Renz, aims to close this gap. The researchers from pediatric radiology and pediatric oncology from 38 university hospitals collect and analyze image data in a structured manner in order to precisely determine the stage of the disease, treatment response and aftercare. The project is part of the Network of University Medicine (NUM) and is funded by the Federal Ministry of Education and Research with around one million euros.

NHL is caused by genetic changes in the lymphocytes. The disease can affect lymph nodes, bone marrow, lungs, liver and spleen. Prof. Renz emphasizes the need for less stressful radiological examinations: "To assess the response, we want to rely even more on radiological examinations, which are less stressful for the children and adolescents affected.”

The aim is to filter information from CT and MRI images in such a way that it can be clearly assigned to a lymphoma stage. This should enable automated image analysis that reliably determines the tumor stage and the risk of relapse. The researchers are using the "NHL-Berlin-Frankfurt-Münster Registry", which has been collecting clinical data from children and adolescents of all NHL subgroups in Germany since 2012.

"Up to now, CT and MRI have not been used to determine a reference standard that radiologists can use as a guide," explains Prof. Renz. One of the main aims of RACOON-RESCUE is therefore to develop a standardized, automated image analysis tool. This new method will not only use radiological image data, but also clinical data to improve the prediction of treatment success and determine the individual risk of relapse.

Artificial intelligence will be used to automatically evaluate a unique data set. Ultimately, a digital diagnostic mask is to be developed that determines tumor stages directly from CT and MRI images, evaluates the response to therapy and improves tumor follow-up care.

MEDICA-tradefair.com; Source: Hannover Medical School