MITK: the DKFZ Medical Imaging Interaction Toolkit

AIHealth

On-premise clinical platform with local LLMs, RAG on FHIR/DICOM data, diagnostic support, remote follow-up. Architecture designed for the MDR pathway.

Discover AIHealth →

Digital Health

Medical software development compliant with CE and MDR regulatory standards. Clinical decision support systems, AI integration in clinical workflows.

Discover →

A European alternative platform

Beyond 3D Slicer — developed in the United States with NA-MIC consortium support — the open source medical imaging landscape includes a European-born platform of comparable maturity: MITK — Medical Imaging Interaction Toolkit — developed at DKFZ (Deutsches Krebsforschungszentrum) in Heidelberg by the Medical and Biological Informatics division, coordinated over the years by Hans-Peter Meinzer and today by Klaus H. Maier-Hein.

MITK was launched in 2003 with initial focus on oncology — liver cancer, surgical intervention planning, tumour mass segmentation — but has grown to cover a broad set of medical imaging use cases. The current version at time of writing is MITK 2016.03, released in March 2016.

The licence is BSD 3-Clause; the code is hosted on a public Git repository and accepts community contributions.

Architecture

MITK is built on a stack of mature open source libraries:

• ITK (Insight Toolkit) — analysis components (segmentation, registration, morphology, filters)
• VTK (Visualization Toolkit) — 3D rendering, visualisation pipeline
• Qt — C++ GUI framework, provides widgets and event loop
• DCMTK — DICOM parsing and I/O
• Poco C++ Libraries — system utilities

On top of this stack, MITK provides its own data model to represent clinical objects (image volumes, segmentations, reference points, surface models, plans) with shared properties and persistence. Views (2D axial/sagittal/coronal, 3D) are synchronised and parameterised on the model.

BlueBerry and plug-in architecture

The Workbench version of MITK uses a BlueBerry framework — architecture akin to the Eclipse Plugin Framework ported into C++ Qt. Clinical functionalities are encapsulated into plug-ins with:

• Declarative contributions to the system — menus, views, toolbars, perspectives
• OSGi-like dependency resolution
• Runtime extensibility — plug-ins loadable without recompilation

For a developer wanting to extend MITK with a specific clinical module — e.g. a volumetric analysis plug-in for a tumour type — BlueBerry provides the infrastructure without touching the core.

Shared components across medical imaging projects are collected in CTK (Common Toolkit Platform for medical imaging), an inter-institutional initiative including contributions from MITK, 3D Slicer, Insight Software Consortium, to promote reuse and interoperability.

Functional modules

MITK Workbench 2016.03 includes modules for several clinical areas:

Segmentation

• Interactive editor with 2D/3D painting tools, thresholding, region growing, smoothing
• Semi-automatic Graph Cut, Region Growing 3D segmentation
• Model-based segmentation

Registration

• Rigid, affine, deformable
• Multimodal registration (e.g. CT + MR, PET + CT)
• Landmark-based for manual correction planning

DICOM and I/O

• DICOM import/export with private tag handling
• Support for multi-volume series, 4D series
• Native formats (Nifti, Nrrd, MHA) for exchange with other toolkits
• Anonymisation pipeline

Oncology

• Tumour volumetry with automatic volume and diameter calculation
• Treatment response analysis (RECIST 1.1 — international criteria for solid tumour response)
• PET/CT fusion for radiotherapy planning
• Seed-based segmentation for moderate-size lesions
• ROI statistics — intensity histograms, texture statistics (basis for radiomics)

Image-guided surgery

• Surgical navigation — instrument tracking, patient registration
• Intervention planning for abdominal, neurosurgical, orthopaedic surgery
• Integration with intraoperative navigation systems

Diffusion analysis

• DTI — diffusion tensor, FA, MD
• Deterministic and probabilistic tractography
• Fibre clustering

3D printing and modelling

• Export to STL formats for 3D printing patient-specific anatomical models
• Prosthesis planning

MITK compared with 3D Slicer

The two platforms have parallel developments and are often compared:

Both are valid choices for research groups; choice depends on the prevailing use case, availability of already-developed modules, GUI and architecture preferences.

Products based on MITK

MITK is the foundation of several spin-off and company commercial products and clinical software:

• Fraunhofer MEVIS — Bremen institute focusing on biomedical imaging, produces certified products based on the MITK/MeVisLab stack
• mint medical (Heidelberg) — radiology oncology products derived from MITK
• Research clinical software — many German and European research groups build specialist interfaces on MITK for specific study protocols

The BSD licence allows commercialisation without copyleft restrictions.

Conformance standards

For certified clinical use, base MITK is not a medical device: it is a development toolkit. Anyone building a certified clinical product on MITK must:

• Follow IEC 62304 for the medical software lifecycle
• Produce end-product-specific technical documentation
• Manage risk (ISO 14971)
• Conduct clinical evaluation
• In Europe, obtain CE mark under MDD 93/42/EEC or the upcoming MDR

The advantage of using an open source base like MITK — inspectable, documented, maintained — is vs. building from scratch, but end-product qualification remains the manufacturer’s full responsibility.

Community and development

The MITK development model includes:

• Public Git repository (github.com/MITK/MITK) with pull request code review
• Mailing lists for developers and users
• Annual MITK workshops hosted at DKFZ, open to external users
• Continuous integration on Linux/Mac/Windows
• Periodic releases every 3-6 months

DKFZ remains the main contributor, but the global community includes contributions from university centres (Leipzig, Munich, Oxford, Cambridge, various US, Japanese, Italian centres).

In the Italian context

In Italy MITK is used in some academic and clinical research contexts:

• Politecnico di Milano — cardiovascular imaging projects, integration with surgical navigation systems
• University of Verona — oncology and radiotherapy
• IRCCS in Piedmont, Emilia-Romagna, Lombardy — study protocols
• Medical informatics spin-offs — companies adopting MITK as base for customised clinical products

Adoption in non-research clinical departments is more limited; Italian hospitals tend to use commercial products (Philips IntelliSpace, GE AW Server, Siemens Syngo) for daily diagnostic workstations.

Outlook

MITK development directions for the coming years include:

• Deep learning integration — DKFZ researchers (the same group developing nnU-Net, set to become the reference for medical segmentation) will integrate CNN-based modules in MITK
• Broader Python scripting — community demand for the scripting flexibility 3D Slicer offers
• Cloud and services — MITK as base for shared cloud imaging platforms
• DICOMweb integration — HTTP APIs to expose MITK data to external viewers and analysers
• Standardised radiomics — MITK is a natural environment for the radiomic feature extraction pipeline, an area in rapid expansion

MITK in 2016 represents a mature, European, well-maintained platform in the open source medical imaging ecosystem, complementary to 3D Slicer and suited to specific oncological and surgical scenarios.

References: MITK (www.mitk.org), MITK 2016.03. Deutsches Krebsforschungszentrum (DKFZ) Heidelberg, Medical and Biological Informatics division. Hans-Peter Meinzer and Klaus H. Maier-Hein as coordinators. BSD 3-Clause licence. ITK, VTK, Qt, CTK. BlueBerry plugin framework.