University Model Image Evaluation (MRI/CT)

Brief Description

With every new MRI or CT examination, an automated pipeline intervenes. Image data is processed immediately upon arrival, evaluated with a current AI model, and incorporated as a structured preliminary report into the radiologists' report list. Critical scans appear prioritized, routine images receive a quick initial assessment. All processing steps are fully documented, ensuring quality and traceability.

Technical Process

As soon as a new MRI or CT series enters the hospital PACS, a DICOM router intercepts the images. They are normalized and anonymized, then passed to a current research model, for example nnU-Net for automatic organ or tumor segmentation or MedSAM for prompt-based multi-organ segmentation. The model delivers preliminary findings that are translated into structured diagnoses and immediately integrated into the radiologists' worklist. All inputs and outputs, including feedback, are stored in an audit-proof manner.

Reference Models:
nnU-Net: https://github.com/MIC-DKFZ/nnUNet
MedSAM: https://github.com/bowang-lab/MedSAM

Benefits of the Automated Pipeline

• Faster Diagnoses: Automatic preliminary findings significantly reduce reporting time
• Higher Quality: AI-supported analysis also detects subtle pathologies
• Prioritization: Critical cases are automatically moved to the top of the worklist
• Continuous Learning: Radiologist feedback continuously improves the model
• Compliance: Complete documentation of all processing steps
• Seamless Integration: Direct connection to existing PACS/RIS systems

Technical Innovation

The pipeline utilizes state-of-the-art deep learning models from university research and makes them immediately available for clinical practice. Through containerized architecture, new model versions can be deployed without downtime.

Revolutionary Medical Imaging Diagnostics: This solution combines cutting-edge university research with clinical practice and sets new standards in radiological diagnostics.