Please note the following instructions:

• Please complete the following sections without changing the formatting.
• The blue text is intended for explanatory purposes only and should be deleted in the final application.
• Sections 7–10 should not exceed 10 pages in total. In these sections, you should describe your research project in detail and explain the steps you plan to take to establish a sustainable collaboration during the requested funding period.
• In addition to the scientific description of the project, the proposal should also address the following questions:
  • What is the national and international relevance of your topic within its research field, and who are the national and international competitors in this area?
  • Do you plan to undertake any further internationalization activities?
  • Will Early Career Researchers (ECRs) be involved and/or supported?
  • Is there a transfer aspect to society, policy, or industry? How will this transfer be implemented?
• The application may be written in German or English.
• The project summary must be provided in both German and English.
• The financial plan must always be submitted in German.

• Applications must be submitted electronically. Please send your application together with the required attachments listed below by email to the UA Ruhr office:

  mercur@uaruhr.de

• All files must be submitted in PDF format.

1. CVs

CVs of all applicants must be submitted as one PDF file (the same CVs that were submitted during the proposal/sketch phase).

2. Financial and Cost Plan

A financial plan must be included that:

• Presents the costs of the proposed project.
• Breaks down and justifies the costs by category (personnel costs and material/operational expenses).
• Uses the template provided by MERCUR.
• Specifies for each university:
  • The total amount of funding required for the entire project.
  • The amount of funding required per year.
• Uses the official DFG personnel funding rates when calculating personnel costs.

3. Applicants with Fixed-Term Contracts

If one of the applicants has a fixed-term employment contract that ends before the planned completion of the project:

• An informal confirmation letter from the chair, institute, or faculty must be attached.
• The letter must confirm continued employment until at least the end of the project.
• Applicants may not request or finance their own positions through MERCUR funding.

Project duration: 12 months

Requested start date: January 1, 2026

Project Vision

TRACE develops a trustworthy, explainable, and robust framework for 3D anomaly detection in clinical imaging, with cranial MRI as the primary use case.

The project combines:

• 3D anomaly detection
• Causal explainability
• Data security
• Clinical validation

to create AI systems that clinicians can trust.

Core Contributions

1. Advanced 3D Anomaly Detection

Building upon:

• PatchCore
• AnomalyDINO
• 3D Transformers
• 3D U-Nets

the project extends state-of-the-art methods to volumetric medical imaging.

2. Causal Concept Bottlenecks

The system introduces expert-supervised concepts that connect detected anomalies to clinically meaningful causes.

Benefits include:

• Improved interpretability
• Better robustness
• Increased transparency

3. Watermark-Aware AI

The project studies how medical image watermarking affects anomaly detection performance and robustness.

TRACE serves as the foundation for two major future initiatives:

1. DFG Research Unit : "Causal Clinical AI"

A collaborative research program focused on:

• Causal inference
• Medical imaging
• Explainable diagnostics

2. ELLIS Unit : "Trustworthy and Secure AI in Healthcare"

A new ELLIS research center integrating:

• Clinical data
• AI innovation
• AI security

WP1 – Robust 3D Outlier Detection (RUB)

• 3D feature extraction
• Density-based anomaly scoring
• Watermark robustness analysis

WP2 – Causal Concept Bottlenecks (TU Dortmund)

• Clinical concept definition
• Causal bottleneck architecture design
• Explainable anomaly detection

WP3 – Clinical Validation (All Partners)

• Dataset preparation
• External validation
• Scientific publication

WP4 – Building Trustworthy Clinical AI at UA Ruhr

• Coordination
• DFG proposal preparation
• ELLIS proposal preparation

TRACE will:

• Advance trustworthy AI in healthcare
• Improve anomaly detection in medical imaging
• Strengthen AI transparency and explainability
• Support future clinical deployment
• Establish UA Ruhr as a leading European center for trustworthy clinical AI

This work also serves as seed funding for future large-scale DFG and ELLIS initiatives.