Project summary

The core elements of medical care for patients are a diagnosis and the therapy based on it. Being able to distinguish between sicker patients and those who are less sick, or between patients with a poorer prognosis and those with a better prognosis, is one of the fundamental medical skills that are of central importance, for example, when it comes to intensifying or modifying therapy strategies.

In cardiology, image data are one of the central tools for diagnosing patients in order to better identify precisely these differences. Traditionally, simple, often measurable values are extracted from the acquired image data, which are then subsequently used in clinical decisions and therapy strategies. MRI images of the heart in particular offer a high density of temporal and spatial information, so that these images in particular offer the potential to derive more from them than traditional measured values.

Artificial intelligence (AI) can be used to detect more from acquired images. Similarly, algorithms can learn image data from patients with other data from those patients, such as diseases, hospitalizations, or times of death associated with that very image data. The better one trains the algorithm on the data, the better it will be at correctly predicting clinical events such as death or hospitalizations.

The observational study BeLOVE, which includes patients with a significantly elevated cardiovascular risk profile, offers the opportunity to evaluate cardiac MRI data to a significant degree and thus develop and validate a prognostic model for risk stratification of cardiology patients.
The development and testing of the reliability of such an AI approach for prognostic assessment of patients is the main goal of this research project.

The Belove project is funded by the German Center for Cardiovascular Research (DZHK) and coordinated by the project leader Djawd Hashemi, MD.

Funding reference

• 81X3100220

Duration of the BELOVE project

• December 2022 - November 2023

Project summary

The goal of the FLOTO project is to establish a federated learning consortium to address a clinical question regarding TAVI prosthesis selections involving 9 hospitals from 7 DZHK sites and Heidelberg University Hospital. Novel concepts of Federated Learning (FL) accomplish a decentralized training of neural networks without the underlying data having to leave the individual sites (data sovereignty remains at the sites; possibility to use reproducibility cohorts that represent a broad spectrum of patient characteristics and recording technique). The project aims to achieve a first milestone in terms of data preparation of over 13,000 available patient records and implementation of a test architecture.

Prof. Dr. Anja Hennemuth is together with Priv.-Doz. Dr. med. Simon Sündermann project leader on the part of the Charité - Universitätsmedizin Berlin and is supported by the scientific assistant of the institute, Nina Krüger. The research program is supported by Jun. Prof. Dr. Sandy Engelhardt from the Heidelberg University Hospital. The German Center for Cardiovascular Research e.v. (DZHK) has been the research sponsor since 2022.

Learn more about the project - Annual Report 2021.

Funding reference

• 81X1100226

Duration of the FLOTO project

•  January 2022 - December 2022

Project summary

The goal of the multi-center project is to investigate the suitability of numerical simulation techniques for simplified and accelerated medical device approval. Verification and validation are among the most critical tasks in the development cycle of cardiovascular implantable devices, leading to recalls that can cost companies millions of dollars, ruin their reputations, and directly affect stock prices. At the same time, clinical safety and performance standards and relevant regulations are becoming increasingly stringent. In silico methods for medical device testing and validation represent a promising opportunity to improve safety, efficacy, time and cost efficiency. However, their integration into the product cycle requires the establishment of agreed protocols, standards and shared resources between device manufacturers, authorities and regulators.
The SIMCor project aims to create a computational platform for in silico development, validation and regulatory approval of cardiovascular implantable devices.

The project, whivh is funded by the European Commission (EU), is partly supported by the project coordinator, Prof. Dr. med. Titus Kühne and the project leader at Charité, Dr. rer. med. Jan Brüning, as well as 11 other project partners. The work package leaders of the institute - Prof. Dr. Anja Hennemuth and Prof. Dr. Leonid Goubergrits - are also working on the SIMCor project and are supported by the research assistants, Nina Krüger and Lars Walczak.

Learn more about the project in the video.

Funding reference

• 101017578

Duration of the SIMCOR project

• January  2021 - December 2023

Project summary

Charité and Fraunhofer - two globally recognized research institutions in the fields of medicine, technology and artificial intelligence - are planning an expanded strategic cooperation with the overarching goal of establishing a joint center for AI & intelligent sensor technology at Charité (working title "CePPH - Center for Participatory Precision Health"). In the planned Charité-Fraunhofer Center "CePPH" at the Berlin site, innovative, digital procedures and technologies for clinical healthcare will be researched, developed and validated under real conditions in a practice-oriented environment. The focus is on the clinical benefit combined with the goal of integrating developed solutions directly into the clinical routine of the Charité. CePPH enables a sustainable way to directly derive from clinical questions to what extent AI can support and lead to an improvement in care. The interaction between clinic and technology enables significantly shorter times between the immediate need for clinical action and the implementation of a technological solution.
The pilot project "CePPH Pilot" is planned for an initial period of two years to gradually achieve the goal.
This includes four core areas:

1. »Intelligent Data Stewards«: Enabling efficient data integration, quality assurance and annotation with reusable and clinically unique semantics of data.
2. »Real Lab AI in Medicine" & "AI Factory«: Supports quality-assured collection and integration of real-world data as well as prospective application of AI solutions in clinical care
3. Sensor-based monitoring & automated therapy control in cardiology: explores a sensor-based method for monitoring and computer-assisted therapy control of patients with cardiovascular diseases
4. Patient-participatory AI-driven knowledge-generating systems approach: develops a technology platform to use prospective, real-world oncology data and establishes a knowledge-generating patient-participatory feedback loop from diagnosis to therapy to follow-up on the model project of adaptive therapy for tumors of the upper aerodigestive tract

Project leader is Prof. Dr. med. Titus Kühne from the German Heart Center of the Charité. The third part of the CeppH pilot project is led by Prof. Dr. Anja Hennemuth, with the support of the institute's research associate, Florian Hellmeier and research assistant, Juliana Franz.
Prof. Dr.-Ing. Leonid Goubergrits is also a contributor to the CeppH pilot project.

Duration of the CeppH-Pilot project

• September 2022 - November 2024

The graduate college BIOQIC

In current clinical practice of medical imaging, diagnostic decisions are often made on the basis of qualitative image markers, leading to uncertainties in diagnosis and long training times. The Research Training Group BIOQIC therefore aims to support PhD students in the field of imaging sciences to explore and further develop biophysically-based quantitative medical imaging and to apply it in clinical pilot studies. BIOQIC mediates inter-institutional and interdisciplinary research and teaching in the working field of imaging sciences between basic research-oriented institutes and clinical institutes of Charité - Universitätsmedizin Berlin and innovative regional companies. The research program focuses on the image-based determination of system-independent, tissue-specific, biophysical parameters, including tissue-mechanical parameters, vascular structures and cross-scale flow characteristics.
The training concept comprises 15 sub-projects within the main topics 'Fluid Transport', 'Tissue Mechanics' and 'Tissue Structures', all of which deal with the quantitative analysis of image markers based on multiscale tissue structures from microscopic to macroscopic properties and their representation in medical cross-sectional images.
The project "BIOQIC - BIOphysical Quantitative Imaging Towards Clinical Diagnosis" is jointly supported by the Charité - Universitätsmedizin Berlin as well as the Humboldt-Universität zu Berlin and the Freie Universität Berlin as host universities. The Technische Universität Berlin, the Leibniz Institute for Molecular Pharmacology and the Physikalisch-Technische Bundesanstalt are also involved. The research program focuses on the image-based determination of system-independent, tissue-specific, biophysical parameters.

Project summary

Due to the thin tissue structure and rapid motion, analysis of heart valves with cardiac imaging is particularly challenging. MRI represents the gold standard for hemodynamic assessment of valve function (Fidock, et al., 2019), whereas anatomic assessment is usually based on CT and transesophageal echocardiography (3D TEE) is used for qualitative assessment of valve motion (Capoulade, Piriou, Serfaty, & Le Tourneau, 2017). In principle, MRI has the potential to assess anatomy, motion, and hemodynamics noninvasively, and it would be highly desirable to allow comprehensive quantitative assessment of heart valves with one modality (Garg, et al., 2019). During surgery, endoscopic imaging is used to visualize the valve in a relaxed and thus deformed state (Falk & Kuntze, 2017). Quantification is usually not performed with these images.

The following goals are pursued in the project

• Enable valve detection in MRI and echocardiography for combined assessment of anatomy and hemodynamics
• Evaluation of image-based assessment of valves and myocardium by experiments with animal hearts
• Evaluation of registration methods for pre- and intraprocedural imaging of the mitral valve apparatus.

The project coordinators are Prof. Dr. Ingolf Sack and Professor and Dr. Tobias Schäffter. Prof. Dr. Anja Hennemuth, Principal Investigator, and Ms. Chiara Manini, research associate at ICM, are working on the sub-research project "Biophysical Multimodal quantitative cardiac valve imaging".
The whole project is funded and supported by the German Research Foundation (DFG).

Funding reference

• 289347353

Duration of the BIOQIC project

April 2022 - April 2023

Project summary

To optimize the quality of treatment of valvular heart disease, decisions are discussed and strategies developed in interdisciplinary and cross-site medical teams. Analysis of image and sensor data plays an important role in selecting the most promising therapy. AI-based methods enable the interpretation of this complex data, but cannot represent the experiential knowledge of a medical team. An interaction of AI-based data analysis with the medical expertise of a team could improve therapy planning and implementation. MINIMAKI is therefore working on AI-based, patient-specific models to rapidly simulate different therapy options, making data and methods accessible to medical staff and patients. Mixed reality concepts are also being developed to help the medical team discuss therapy options. The concepts and software solutions developed to integrate AI into data analysis optimize the selection of a therapy for valvular heart disease. Subsequently, innovative interaction concepts with mixed reality help the medical team to overcome difficulties in discussing, planning and performing the complex surgical procedures. The combination of these measures significantly improves the quality of treatment.

The motivation

To optimize the quality of treatment of valvular heart disease, decisions are discussed and strategies developed in interdisciplinary and multisite teams of physicians. Analysis of image and sensor data plays an important role in selecting the most promising therapy. AI-based methods enable the interpretation of these complex data, but cannot represent the experiential knowledge of a team of physicians.

The MINIMAKI project is funded by the German Federal Ministry of Education and Research (BMBF) and supported by the project coordinator Prof. Dr.-Ing. Anja Hennemuth and the project manager Dipl.-Inf. Markus Hüllebrand,. Mr. Priv.-Doz. Dr. med. Simon Sündermann and Juniorprof. Dr. Susanne Michl are also project leaders from the Charité - Universtitätsmedizin Berlin. In addition to the Fraunhofer Institute for Digital Medicine MEVIS, there are other project members of the Deutsches Herzzentrum der Charité that working on the research project.

Announcement 2019

• Adaptive Technologien für die Gesellschaft – Intelligentes Zusammenwirken von Mensch und KI

Funding reference

• 16SV8649

Duration of the MINIMAKI Project

• March 2021 - March 2024

Project summary

Multifrequency magnetic resonance elastography (mMRE) measures the dispersion of viscoelasticity in soft tissues, making it sensitive to extracellular matrix (ECM) and cellular micromechanical interactions. In the first funding phase, we developed mMRE techniques operating at different magnetic field strengths from 0.5 to 7 Tesla to cover a wide range of mechanical excitation frequencies for a comprehensive analysis of viscoelastic properties of normal and diseased biological tissues. Using this innovative multimodal mMRE technology, we have shown that degradation of the ECM due to inflammatory processes or accumulation of fibrous proteins is accompanied by significant changes in the internal mechanical friction of soft tissue. Mechanical friction leading to viscoelastic dispersion thus provides a sensitive probe for mMRE-based diagnosis beyond fibrosis-related tissue stiffening. Initial clinical trials quantifying dispersion by mMRE have been performed in patients with inflammatory bowel disease, nephritis, pancreatitis, hepatitis, liver fibrosis, and cancer. For the next funding phase, we plan to further increase the sensitivity of MRE to viscous dispersion by extending the dynamic range of mMRE to ultra-low and ultra-high frequencies. We and others have provided initial evidence that ultralow frequencies are useful in MR.

The CRC 1340 project of the Charite - Universitätsmedizin Berlin is supervised by the project coordinator Prof. Dr. Bernd Hamm and Prof. Dr. Ingolf Sack.
Furthermore, Prof. Dr. Anja Hennemuth is working as project leader on the research project, with active support from the scientific assistant of the ICM, Heloise Bustin.

Funding for the research project is provided by the German Research Foundation (DFG) and has been supported since 2018.

Funding reference

• 372486779

Duration of the SFB 1340 project

• July 2022 - June 2026

Project summary

Start point

Cardiovascular disease (CVD) is the leading cause of death worldwide. The personal profile of “classic” common risk factors like type 2 diabetes, hypertension, and dyslipidaemia are used by algorithms for individual CVD lifetime risk prediction and prophylaxis decisions for populations without established cardiovascular disease. One particular very high-risk group are subjects that already had an acute cardiovascular event (CVE) like acute coronary syndrome (ACS), acute heart failure (AHF) or acute stroke. In European populations, these group of patients have an estimated 10-year risk >10% for cardiovascular death and >20% for recurrent events. However, recent studies suggest that risk among these patients may substantially vary from <10% to >30%. Survivors of cardiovascular events often experience recurrences of the same disease but are also frequently affected by other vascular disease entities. Not surprisingly, an optimized therapy for one CVD condition can reduce the risk of suffering from several vascular disease manifestations. An effective cardiovascular secondary prevention requires accurate risk prediction. Compared to primary prevention data to develop risk prediction models for patients with established CVD are scarce. Currently, mostly standard measures for major risk factors like blood pressure, HBA1c or LDL-cholesterol are used in clinical practice for risk prediction after a cardiovascular event with several limitations.
New biomarkers could play a pivotal role for risk prediction and treatment adjustment. Such biomarkers and risk prediction models will also have to take complex pathobiological interactions into account. A recent study showed a substantial variability in the long-term risk for recurrence and death after an acute CVD event among patients. The risk remains very high in a significant number of patients even in those with optimal secondary prevention therapy. One reason for this finding could be the increasing multimorbidity in patients with CVD. While the overall age-adjusted incidence and mortality of CVD has decreased over the last decades, the proportion of incident CVD patients with multiple cardio-metabolic as well as and non-cardiac comorbidities markedly increased. It is also well known, that acute events commonly have effects on remote or distant organs that might not seem to be involved in the first place. Examples are associations between heart failure and kidney function, cardiac complications after stroke or interaction with other organs after stroke. Organ crosstalk mechanisms have a strong impact on short-term outcome of acute CVD while their effect on the long-term course is much less understood.
In research on risk prediction and secondary prevention strategies such factors have not yet received sufficient attention. Furthermore, the identification of common pathological mechanisms underlying vascular dysfunction of organs is of key importance to establish new effective intervention options. Given these complexities, it should be considered the heterogeneity and individual variability in the phenotypes as well as in the course of CVD. A classic reductionist approach to presume one single common phenotype to be representative of a group of persons affected by a vascular disease may restrict the understanding of the mechanism of disease as well as the potential for more individually tailored treatments. Research strategies proposed to promote personalized medicine are promising methods to overcome these limitations (see figure 1). Principles of these approaches are the identification of biomarkers as individual risk predictors and the mechanism-based characterization of subgroups of patients that will enable the development of more specific, personalized treatments. Key methods used are deep phenotyping to measure all aspects of disease manifestations and which is to be complemented by omics technologies and “mining” of clinical and research data using artificial intelligence.

The project Image-based personalized prediction of residual risk and prognosis of cardio/cerebrovascular diseases based on the mapping data of the BELOVE study is funded by the German Research Foundation (DFG) in the SPP 2177: Radiomics: Next Generation of Medical Imaging from 2023 for 3 years. Professors Prof. Dr. Anja Hennemuth, Univ. Prof. Dr. Jeanette Schulz-Menger and Dr. Kersten Villringer jointly lead and supervise the research project. Another great support is the work of the research assistant of the ICM, Mrs. Ann Laube.

Funding reference

• 515294457

Duration of the Radiomics project

• May 2023 - April 2024

Project summary

HFpeF is a heterogeneous syndrome associated with increased myocardial stiffness, fibrosis, fat accumulation, microvascular dysfunction, abnormal ventricular-aortic coupling, and deregulation of cardiac energy metabolism in. These phenotypes of cardiac remodeling may depend on the predisposition, intensity, and duration of the underlying triggers. Cardiovascular magnetic resonance (CMR) imaging is the current gold standard for assessing cardiac size and function as well as myocardial tissue changes such as fibrosis, inflammatory response, and fatty infiltration. New CMR approaches allow quantification of luminal blood flow, microvascular perfusion, and blood oxygenation; quantification of epicardial and myocardial fat; and functional reserve by performing an MR-compatible physiological stress test. While conventional analysis methods use only a fraction of the data provided by CMR, artificial intelligence (AI)-based image analysis allows for more comprehensive consideration of complementary CMR sequences and parallel assessment of other organ systems (kidney, lung). Our central hypothesis is that innovative CMR applications combined with AI will enable imaging-based differentiation of prevalent HFpEF cardiac pathologies.

Projekt B06

The prevalence of heart failure (HF) is increasing, and morbidity and mortality are unacceptably high. The diagnosis and clinical management of HF are currently guided by left ventricular ejection fraction, because HF with preserved ejection fraction (HFpEF) is considered a different disease from HF with reduced ejection fraction (HFrEF). Treatment strategies have been developed for HFrEF to improve prognosis, but most have proven ineffective in HFpEF, so we are still unable to provide specific therapies for this large subset of HF patients (~50%). In this CRC, we are taking an interdisciplinary approach, from organism to organ to cell to molecule, to characterize HFpEF as a systemic and heterogeneous disorder. Our central hypothesis is that dysregulation of systemic hemodynamic, metabolic, and inflammatory pathways contributes to distinct HFpEF phenotypes with specific pathophysiological features that respond differently to targeted therapies. By leveraging our expertise in multi-omics, advanced imaging, functional phenotypic analysis, AI, and modeling, we will provide a better foundation for a comprehensive mechanistic understanding of the disease to develop innovative therapies for individual patients.

We have focused our research program on defined mechanical, metabolic, inflammatory, and immunological triggers, their respective downstream signaling pathways, and specific cardiac response patterns. We bring together young and established basic scientists and clinicians with expertise in translational cardiology, functional genomics, cell and molecular biology, systems medicine, proteomics, metabolomics, and bioinformatics to achieve improved in-depth phenotyping and classification of HFpEF in relevant animal models and patients as a basis for individualized therapy.

We emphasize training the next generation of heart failure researchers, focusing on good scientific practice, equal opportunity, patient engagement, and the principles of 3Rs to foster a culture of collaboration, communication, and cross-thematic activity. Together, we will lay the foundation for improved mechanistic understanding and classification of HFpEF that combines advanced imaging, multi-omics, and classical risk factor analysis to provide the basis for causal, individualized therapies. We believe that our approach will be instrumental in improving the treatment of HFpEF as one of the greatest unmet clinical needs in medicine.

The German Research Foundation (DFG) has been funding the project SFB 1470 since the beginning of 2022. In addition, the Deutsche Herzzentrum der Charité and the Charité University Medicine Berlin are participating in the research program alongside our Institute for Cardiovascular Computer-Assisted Medicine. The project is coordinated by Prof. Dr. Anja Hennemuth and Prof. Dr. Sebastian Kelle.
The postdoc and research associate, Dr. rer. nat. Lars Walczak, is working on the SFB1340 B06 project.

Funding reference

• 437531118

Duration of the SFB1470 B06 project

• January 2022 - December 2022

Project summary

Myocarditis is a cause of severe heart failure in children, with severe disease courses especially in young children. Observation of data from the prospective multicenter myocarditis registry for children and adolescents "MYKKE" leads to the hypothesis of different pathomechanisms in these cohorts, which could also be due to different immunological or genetic responses. Unfortunately, this has not been well studied. Therefore, the concept of this project is to provide risk stratification and disease progression analysis in pediatric suspected myocarditis by developing an artificial intelligence (AI) technique based on imaging data and statistical analysis. The primary goal is to define patient groups with different phenotypes using machine learning in image analysis and to leverage collaboration with the Department of Medical Statistics to integrate image data into an AI-based statistical analysis for further risk prediction. Prof. Friede is also involved in further statistical analysis of the clinical data of the studied patient cohort and is familiar with the clinical dataset.

The project program Shared Expertise MYKKE is funded by the research organization "Deutsches Zentrum für Herz-Kreislauf-Forschung e.v." since 2021. The subproject leaders Prof. Dr. Anja Hennemuth and Dr. med. Franzsiksa Seidel and the research associate, Ms. Léa Ter-Minassian, are working on the Shared Expertise MYKKE concept on behalf of the Deutschen Herzzentrums der Charite (DHZC). The University Medical Center Göttingen is represented in the research project by the Institute Director for Medical Statistics, Prof. Dr. Tim Friede.

Studies 2021

• Studie findet keine Hinweise auf Herzmuskelentzündung nach COVID-19-Erkrankung bei Kindern
• Gendefekte könnten Risiko für schwere Herzmuskelentzündungen im Kindesalter erhöhen

Funding reference

• 81X2100274

Duration of the MINIMAKI Project

• April 2022 - March 2024