Rijhwani et al., BVM 2026 — An interdisciplinary toolbox that bridges radiology and pathology imaging within a unified DICOM-based environment, deployable as a standalone tool or as a Kaapana extension.
.jpg)
The framework addresses two fundamental requirements for multimodal image evaluation:
- Modality mapping — Whole slide images (WSIs) are converted to DICOM format using wsi-dicomizer and spatially linked to their corresponding radiology images by resolving the anatomical region via TotalSegmentator.
- Synchronized visualization — A split-screen web viewer embeds OHIF (radiology) and SLIM (pathology) in a single coherent interface, with interactive overlays that link directly to the corresponding pathology slide.
All data is managed through a dcm4chee PACS using standardized DICOM tags, making the pipeline fully reproducible and interoperable with existing medical imaging infrastructures.
- Split-screen viewer — OHIF (radiology) and SLIM (pathology) unified in a single web interface
- Automated multimodal mapping — TotalSegmentator derives organ context from WSIs and segments corresponding structures in radiology volumes
- DICOM-native — Both modalities handled through dcm4chee PACS; WSI conversion via wsi-dicomizer
- Interactive overlays — Radiology overlays link directly to the matched pathology slide
- Kaapana integration — Available as a Kaapana extension with an Airflow-orchestrated auto-segmentation pipeline
- Standalone deployment — Can be run independently via Docker Compose without a full Kaapana installation
combinedmodalityviewer/
├── Application_folder/ # Split-screen viewer application (OHIF + SLIM + nginx + dcm4chee)
│ └── SViewer2.0/
│ ├── split-viewer/ # Vite/React wrapper combining both viewers
│ └── Viewers/ # OHIF Viewers fork with DKFZ customizations
├── Docker_Images/ # Pre-built Docker image tarballs and deployment notes
├── Kaapana_Extension/ # Kaapana workflow extension (Helm chart + Airflow DAG)
│ └── Auto-segmentation-pipeline/
│ └── processing-containers/ # TotalSegmentator-based organ segmentation container
├── Readme/ # Demo video, screenshots, flowchart, and associated paper
├── Report/ # Full technical report (LaTeX source + compiled PDF)
└── LICENSE.txt
Two deployment paths are available: build from source (recommended for development) or use pre-built Docker images (quickest setup).
Requirements: Docker, Docker Compose, Git
1. Clone this repository
git clone https://github.com/MIC-DKFZ/combinedmodalityviewer.git
cd combinedmodalityviewer2. Clone and start the SLIM pathology viewer
The SLIM viewer must be cloned and started separately before the main stack:
git clone https://github.com/ImagingDataCommons/slim.git
cd slim
docker compose up -d --no-deps app
cd ..3. Start the main application stack
Navigate to the Nginx + dcm4chee recipe and bring up all services:
cd Application_folder/SViewer2.0/Viewers/platform/app/.recipes/Nginx-Dcm4chee
docker compose upThis starts the OHIF viewer, split-screen wrapper, split-server, dcm4chee PACS (LDAP + PostgreSQL + archive), and nginx reverse proxy. Once running, open http://localhost in your browser.
Requirements: Docker, Docker Compose
1. Download the pre-built image tarball
Download nginx-SPLITVIEWER-images.tar from hub.dkfz.de/f/108118972 and load it:
docker load -i nginx-SPLITVIEWER-images.tar2. Start the application stack
cd Application_folder/SViewer2.0/Viewers/platform/app/.recipes/Nginx-Dcm4chee
docker compose up -dOpen http://localhost in your browser.
Note: The pre-built images are built for x86_64. If you need to migrate existing Postgres or LDAP data, back up and restore the respective Docker volumes before starting.
For institutional deployments on a MicroK8s/Kubernetes cluster, the full Kaapana-based platform (including the automated organ-segmentation pipeline) can be deployed via the provided Helm script:
cd Application_folder/SViewer2.0
bash deploy_platform_0.5.2.shThe script runs preflight checks, pulls the Kaapana admin chart from the HZDR registry, handles version migrations, and installs all platform components.
For detailed architecture, configuration options, and evaluation, see the technical report in Report/main.pdf.
This work was presented at Bildverarbeitung für die Medizin (BVM) 2026:
Nilesh P. Rijhwani, Titus J. Brinker, Neher Peter, Nolden Marco, Klaus Maier-Hein, Christoph Wies, Maximilian Fischer. Bridging Radiology and Pathology: A DICOM-based Framework for Multimodal Mapping and Integrated Visualization. In: Handels H. et al. (Hrsg.), Bildverarbeitung für die Medizin 2026, Informatik aktuell, Springer Fachmedien Wiesbaden, 2026. https://doi.org/10.1007/978-3-658-51100-5_62
The paper PDF is included in Readme/978-3-658-51100-5_62.pdf.
If you use this framework in your research, please cite:
@InProceedings{rijhwani2026bridging,
author = {Rijhwani, Nilesh P. and Brinker, Titus J. and Peter, Neher
and Nolden, Marco and Maier-Hein, Klaus and Wies, Christoph
and Fischer, Maximilian},
title = {Bridging Radiology and Pathology: {A} {DICOM}-based Framework
for Multimodal Mapping and Integrated Visualization},
booktitle = {Bildverarbeitung f{\"u}r die Medizin 2026},
editor = {Handels, H. and others},
publisher = {Springer Fachmedien Wiesbaden GmbH},
year = {2026},
doi = {10.1007/978-3-658-51100-5_62},
}Copyright © 2025 Division of Medical Image Computing, German Cancer Research Center (DKFZ), Heidelberg, Germany.
This project is licensed under the Apache License 2.0 — see LICENSE.txt for the full text.
This framework integrates several open-source components, each governed by its own license. Users must comply with the individual licenses of all incorporated tools, including but not limited to:
| Component | License |
|---|---|
| OHIF Viewers | MIT |
| SLIM | Apache 2.0 |
| dcm4chee | LGPL 2.1 |
| TotalSegmentator | Apache 2.0 |
| wsi-dicomizer | Apache 2.0 |
| Kaapana | Apache 2.0 |
The respective license texts are distributed with each component. By using this software, you agree to abide by all applicable third-party licenses in addition to the Apache 2.0 license of this repository.