Networks & Platforms

We actively contribute to national and international research networks and platforms, fostering collaboration in AI and medical imaging. Our focus is on scalable data analysis, federated learning, and robust research software to drive innovation. By integrating cutting-edge methods into clinical and scientific workflows, we enhance AI’s real-world impact.

Kaapana

Kaapana is an open-source toolkit for state-of-the-art platform provisioning in the field of medical data analysis. The applications comprise AI-based workflows and federated learning scenarios with a focus on radiological and radiotherapeutic imaging.

Obtaining large amounts of medical data necessary for developing and training modern machine learning methods is an extremely challenging effort that often fails in a multi-center setting, e.g. due to technical, organizational and legal hurdles. A federated approach where the data remains under the authority of the individual institutions and is only processed on-site is ideally suited to overcome these difficulties.

Kaapana provides a framework and a set of tools for sharing data processing algorithms, for standardized workflow design and execution as well as for performing distributed method development. This facilitates data analysis in a compliant way enabling researchers and clinicians to perform large-scale multi-center studies.

By adhering to established standards and by adopting widely used open technologies for private cloud development and containerized data processing, Kaapana integrates seamlessly with the existing clinical IT infrastructure, such as the Picture Archiving and Communication System (PACS), and ensures modularity and easy extensibility.

Contact: Ünal Akünal, Philipp Schader
Links: https://www.kaapana.ai/; https://github.com/kaapana/kaapana

RACOON – RAdiological COOperative Network

The RACOON initiative establishes a nation-wide infrastructure for the structured acquisition, processing and analysis of radiological imaging data. By connecting all 38 university clinics providing images including structured reports of the diagnosis, it is able to form a solid foundation for radiological research in Germany. The established infrastructure and collected data sets are of use for early detection systems and AI supported medical decision support systems and therefore form a solid foundation towards archiving pandemic preparedness.
The department of medical image computing provides its expertise in building federated machine learning infrastructures. It contributes the Kaapana software platform allowing easy cohort definition, centralized and federated machine learning for image analysis. Furthermore, method sharing between the partners is streamlined by providing supporting execution of containerized methods either on-site as part of the local RACOON-Nodes or in the central component as part of RACOON-Central.

Contact: Peter Neher

Helmholtz Metadata Collaboration (HMC) - Hub Health

The Helmholtz Metadata Collaboration Platform develops concepts and technologies for efficient and interdisciplinary metadata management spanning the Helmholtz research areas Energy, Earth and Environment, Health, Matter, Information, Aeronautics, Space and Transport. As HMC Hub Health, we support researchers and clinicians in structuring, standardizing, and expanding the collection of metadata to facilitate the re-use, interoperability, reproducibility, and transparency of their data.

Contact: Marco Nolden, Lukas Kulla
Link: https://www.helmholtz-metadaten.de

Joint Imaging Platform (JIP)

The Joint Imaging Platform ( JIP ) is a strategic initiative within the German Cancer Consortium (DKTK). The aim is to establish a technical infrastructure that enables modern and distributed imaging research within the consortium. The main focus is on the use of modern machine learning methods in medical image processing. It will strengthen collaborations between the participating clinical sites and support multicenter trails.

The project attempts to address the organizational challenges of data protection requirements by exchanging and distributing the processing methods rather than patient data.

Contact: Peter Neher

Helmholtz Imaging - Applied Computer Vision Lab

The Applied Computer Vision Lab, operating within the Helmholtz Imaging framework, is dedicated to catalyzing research in Helmholtz and beyond. The lab specializes in providing customized image analysis solutions and building tailored AI algorithms to address specific challenges. In this context, the lab actively engages in Helmholtz Imaging Collaborations, leveraging their expertise to address imaging-related challenges in partnership with other researchers in the Helmholtz Association. On a broader scale, they leverage their experience to develop and provide out-of-the-box solutions like nnU-Net which are applicable across domains and catalyze the development of new algorithms. Moreover, the lab is committed to enhancing algorithm evaluation by organizing competitions, assisting in the selection of appropriate metrics, and supporting evaluation schemes. Currently, the lab is collaboratively working to expand nnU-Net, aiming to cover a broader range of imaging domains and tasks, such as pixel-wise regression and instance segmentation. Their dedication extends to creating foundation models with the aim of revolutionizing the development of state-of-the-art AI methods across diverse imaging domains.

Contact: Fabian Isensee
Links: https://helmholtz-imaging.de/

Effective Privacy-Preserving Adaptation of Foundation Models for Medical Tasks

Foundation models in the vision domain are large machine learning models pre-trained on vast amounts of natural images to extract relevant features from their input data. In the PAFMIM project, we aim at adapting existing foundation models to sensitive medical images, such as CT and MRI data. The challenges include ensuring a good performance of the models on medical images and guaranteeing privacy for the medical images. Our project addresses the challenges in a joint approach combining competencies from the perspective of medical image computing and privacy-preserving machine learning.

Contact: Santhosh Parampottupadam
Links: https://pafmim.github.io/

Secure Decentralized Medical Image Analysis

The aim is to enhance security in decentralized medical image analysis by developing strategies that leverage the usage of data from various sources while complying with strict privacy laws. It focuses on evaluating existing practices and developing new, secure algorithms for medical image analysis. It includes practical testing in clinical environments to enhance the technique's effectiveness and security. The aim is to enable safe and improved medical imaging analysis for widespread use in national healthcare studies. Within the project, the RACOON initiative plays a crucial role, providing the framework and infrastructure for implementing and testing these advanced security strategies in medical imaging across various clinical settings nationwide.

Contact: Benjamin Hamm

CCE-DART

The CCE-DART (CCE Building Data Rich Clinical Trials) project, funded by the European Union, focuses on improving the efficiency and effectiveness of clinical trials in oncology. It is carried out by Cancer Core Europe (CCE), a consortium of seven comprehensive cancer centers within Europe.

At the German Cancer Research Center (DKFZ), a multidisciplinary team from five different departments is actively working towards this goal. One of the contributors is the Department of Medical Image Computing, which is specialised in building federated image analysis infrastructure. We are building a data sharing and analysis platform based on the Kaapana framework. This platform will enable researchers to access relevant imaging data and perform federated image analysis more efficiently.

Contact: Philipp Schader

M²OLIE (“Mannheim Molecular Intervention Environment“)

M²OLIE is one of nine Research Campuses in Germany that have been funded by the Federal Ministry of Education and Research since 2012 as part of the “Research Campus – Public-Private Partnership for Innovation” Initiative. In the sub-project SIM²BA (Standardization & Interoperability of MultiModal Image Analysis Methods) we investigate methods to connect data and machine learning methods contributed by different partners to make them available for evaluation in the Closed Loop process of the research campus.

Contact: Maximilian Fischer

Quality-controlled analysis of large population-based imaging datasets

In this project, we explore approaches for conducting quality controlled image analysis in large health studies like the German National Cohort or UK Biobank. We specifically focus on efficient and reliable techniques for performing quality control on machine-generated image segmentations in the absence of ground truth, as a prerequisite for ensuring the correctness of image-based measurements. To establish time and annotation efficient solutions that are largely applicable on various organs of interest, we investigate the use of uncertainty quantification methods as an efficient means for estimating error. We closely examine goal-specific performance measures to assess how well error estimators are suited for accomplishing important tasks of segmentation quality control. Beyond this, we focus on the development of DL based techniques for detecting and repairing notorious MR imaging artifacts that otherwise require tedious manual effort to handle, and that are known to corrupt measurements derived from affected imaging data.

Contact: Tobias Norajitra