Applied Medical Image & Data Analysis Lab
AMIDL
Welcome to the website of our research group
Lab Maros & Wenz
Introduction & Mission statement
In medicine we generate vast amounts of data on a daily basis. However, the manner in which it is obtained is, for the most part, unstructured and therefore rarely ready to be utilized to retrieve valuable insights or synthesize new knowledge. Text and Images are the two classical outputs/pillars of diagnostic medical (especially radiological) data. Furthermore, high-throughput genomic assay technologies are becoming indispensable in state-of-the-art diagnostic and therapeutic decision-making.
The objective of our new research group is to combine these facets of information and exploit the recent advancement and renaissance of machine learning and deep learning algorithms with the aim to improve the accuracy and depth of radiological diagnostic processes as well as the quality and information content of our reports.
Through such efforts, our ultimate goal (and hope) is to work towards safer personalized and more effective patient care while strengthening the role of radiologists in the future.
Recent publications & preprints
Recent Event>
Rainstorms with a splash of blood: machine learning-based predictive analytics for hemorrhagic stroke admissions based on weather systems
Poster presentation, P0165 Epidemiology & Risk Factors, 2024-05-16
Recent Event>
LLMs for radiological report summarization
Oral presentation, GeMTex LLM Workshop, Leipzig, Germany, 2024-05-14
Past Events (2017-2024)
> ESOC 2024, Basel, Switzerland
Rainstorms with a splash of blood: machine learning-based predictive analytics for hemorrhagic stroke admissions based on weather systems
P0165, Epidemiology & Risk Factors, 2024-05-16
- Local results from the WEather-based STroke event and Outcome Risk Modeling (WE-STORM) project (PI Dr. Maros).
> ASNR 2023, Chicago, IL, USA
Predictive modeling of acute stroke admissions based on meteorological weather systems to improve clinical resource allocation
Oral presentation, Annual Meeting of the American Society of Neuroradiology (ASNR) , 04/29-05/03/2023
Predicting acute stroke occurrence related to weather systems to improve clinical resource allocation: a machine learning approach.
RPS 2105, AI in stroke and neurovascular imaging, 2023-03-04 16:00 - 17:30
- Local results from the WEather-based STroke event and Outcome Risk Modeling (WE-STORM) project (PI Dr. Maros).
> DGNR 2021, Kassel - 56. Jahrestagung der Deutschen Gesellschaft für Neuroradiologie e.V.
- Machine learning on top of deep learning-based brain volumetry segmentation to support neuroradiologists in diagnosing neurodegenerative disorders
State of the Art - Neurodegeneration, Best-of-Abstract Power Pitch, 10:10-10:15, 10/8/2021
> DGNR 2019, Frankfurt - 54. Jahrestagung der Deutschen Gesellschaft für Neuroradiologie e.V.
The hidden prowess of NLP and radiological reports: a text mining pipeline for machine learning-assisted diagnostics and reporting
Oral presentation. Forschung: Zerebrovaskulär und anderes, 8:30-10:00, 10/12/2019
How to break the bottleneck of AI driven radiology and get your hands on the new gold: "RIPE" a solution for data extraction and analysis pipeline
Poster Presentation. Posterbegehungen IX (KI & Neuroonkologie), 8:30-9:30, 10/12/2019
**Breaking news: the work received the poster prize of the KI & Neuroonkologie poster session.**
In cooperation with the Experimental Neuroradiology and preclinical Imaging Group/AG
Longitudinal in vivo evaluation of large vessel vasospasm in mice using a micro-CT
Poster Presentation. Posterbegehung II (Interventional II), 14:15 - 15:15, 10/10/2019
> ASNR 2019, Boston, MA, USA - 57th Annual Meeting of the American Society of Neuroradiology (ASNR)
Deep learning-based imaging classifier for improving differential diagnosis between primary and metastatic brain tumors with highly overlapping MRI morphologies
Quality assessment of structured multi-parametric MRI reports of the prostate based on RADLEX mapping of urosurgical key information content
> DGNR 2018, Frankfurt am Main, Germany
Deep learning-based imaging classifier for improving differential diagnosis between primary and metastatic brain tumors
> ASNR 2018, Vancouver, BC, Canada - 56th Annual Meeting of the American Society of Neuroradiology (ASNR)
Real-time decision support for cerebral vasospasm detection on conventional angiograms using deep learning
> ECR 2018, Vienna, Austria - European Congress of Radiology
Structured reporting supports junior readers and improves PI-RADS conformity of multi-parametric MRI reports of the prostate-based on cross-lingual RADLEX annotations
Opportunities in automatically annotating radiology reports with RadLex terms
> DGNR 2017, Cologne, Germany - 52nd Annual Meeting of the German Society for Neuroradiology (DGNR)
Online Webtool für objektive Vergleichbarkeit klassischer Befundtexte und semiautomatisch angefertigter strukturierter Befunde;
This work received the Poster Prize in Computer Science section @ 52nd Annual Meeting of the German Society for Neuroradiology (DGNR)
Academic | Federal Cooperations & Fundings
The German Federal Ministry for Economic Affairs and Energy ( BMWi ) funds innovative research projects through its ( Zentrales Innovationsprogramm Mittelstand, ZIM ) project to foster cooperation between academia and industry. We have successfully finished working on such a BMWi-ZIM Project in partnership with EMPOLIS GmbH (see by <Projects> ).
is a joint (consortial) project including eight German universities that have agreed to share data to enhance patient-centered research and clinical care. It is funded by the German Ministry for Education and Research (BMBF) within the framework of the Medical Informatics Initiative (MI-I, 2015) to foster IT innovations for healthcare research and medical care. Our research group is involved in the MIRACUM Network with regard to Neurovascular and Neurooncological Use Cases.
BMBF - Medical Informatics Initiative (MII)
MIRACUM Consortium - Junior Research Group - <MIDorAI>
The Federal Ministry of Education and Research (BMBF) is promoting 21 junior research groups with around 30 million euros (2020-2026) at the interface of computer science and medicine within the framework of the Medical Informatics Initiative (MII) as a foundation for newly established medical informatics professorships.
MIRACUM the largest MII consortium has taken on the supervision of six junior research groups with impressive research focuses in this regard.
One of which is headed by Dr. Maros since 10/2022:
Medical Informatics for Holistic Disease Models in Personalized and Preventive Medicine (MIDorAI).
Industrial Partners & Cooperations
EMPOLIS has long-lasting and extensive experience in information management. We have a working cooperation with the Healthcare Analytics Branch of EMPOLIS with Dr. Benedikt Kämpgen (Head of SU Healthcare). Our cooperation has recently been awarded federal funding by the BMWi's ZIM (see above) for a two-year time span to develop software solutions for improving the radiological reporting process. We are currently focusing on identifying automatable objective quality measures for reporting.
Institutional Affiliations
Department of Neuroradiology
University Medical Center
Heidelberg University