Fraunhofer Institute for Digital Medicine MEVIS
Fraunhofer Institute for Digital Medicine MEVIS

Radiomics discovers relationships between image features and clinical data. We apply techniques from statistics and machine learning to build predictive models for patient outcomes such as survival, therapy response, or side effects. The images are either characterized with engineered features or are directly fed into a neural network.

Radiomics discovers relationships between image features and clinical data. We apply techniques from statistics and machine learning to build predictive models for patient outcomes such as survival, therapy response, or side effects. The images are either characterized with engineered features or are directly fed into a neural network.

Radiomics discovers relationships between image features and clinical data. We apply techniques from statistics and machine learning to build predictive models for patient outcomes such as survival, therapy response, or side effects. The images are either characterized with engineered features or are directly fed into a neural network.

Radiomics discovers relationships between image features and clinical data. We apply techniques from statistics and machine learning to build predictive models for patient outcomes such as survival, therapy response, or side effects. The images are either characterized with engineered features or are directly fed into a neural network.

Radiomics discovers relationships between image features and clinical data. We apply techniques from statistics and machine learning to build predictive models for patient outcomes such as survival, therapy response, or side effects. The images are either characterized with engineered features or are directly fed into a neural network.

Radiomics

 

© Fraunhofer MEVIS

International Radiomics Platform (SATORI)

The International Radiomics Platform (SATORI) was developed in a public-private partnership with the German and Austrian Societies of Radiology (DRG/ÖRG). It is a web-based platform for collecting, sharing, and annotating image and clinical data. Custom neural networks can be integrated. In a first study, it was used for myocardial segmentation and detection of late gadolinium enhancement.

 

DFG Radiomics Melanoma

We participate in the DFG Priority Program SPP 2177 "Radiomics: Next Generation of Biomedical Imaging", which aims to advance the diagnostic and prognostic value of medical imaging through radiomics. In our subproject with the University Hospital Tübingen, we investigate whether radiomic analysis of whole-body tumor load can predict whether melanoma patients will respond to a given therapy.

 

© Fraunhofer MEVIS

PANTHER

In the PANTHER project, we cooperated with Siemens Healthineers and the University Hospital Munich to use CT images more effectively in tumor follow-ups. Focusing on liver metastases in colorectal and pancreatic cancer and spleen involvement in lymphoma, we explored shape characteristics and distribution patterns that are predictive of patient survival or response to treatment. Our results may help physicians make better informed therapy decisions.

RESECT

The goal of the RESECT project is to predict the resectability of a prostate cancer based on MR images. A broad consortium with two university hospitals, public-private partnerships, an established company, and the German Röntgen Society will create a translational value chain and strive to develop a solution that is ready for clinical use.

End-Stage Liver Disease

In a close collaboration with Brigham and Women's Hospital in Boston established by former institute director Ron Kikinis, we aim at finding new objective criteria to decide which patients should get a liver transplant first. This is crucial due to the limited availability of transplants and the low life expectancy with advanced cirrhosis. In a first study, we compared the predictive value of radiomic features from MRI with established criteria based on lab values.