# MIC-DKFZ/medicaldetectiontoolkit

The Medical Detection Toolkit contains 2D + 3D implementations of prevalent object detectors such as Mask R-CNN, Retina Net, Retina U-Net, as well as a training and inference framework focused on dealing with medical images.

Repository: https://github.com/MIC-DKFZ/medicaldetectiontoolkit
Canonical: https://ross.abutalabs.com/products/medicaldetectiontoolkit
Language: Python
License: Apache-2.0
License Family: permissive
Topics: detection, retina-unet, object-detection, mask-rcnn, 3d-models, deep-learning, retina-net, deep-neural-networks, segmentation, semantic-segmentation, u-net, 3d-mask-rcnn, 3d-object-detection, medical-imaging, medical-image-processing, medical-image-computing, medical-image-analysis, pytorch-deeplearning, pytorch-cnn, pytorch-implementation
Last push: 2024-06-17T22:47:46+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2882, "days_push": 807, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1357, forks 294 (observed 2026-08-28T04:04:29.525058+00:00)

## What it is
A PyTorch framework providing 2D and 3D implementations of object detectors like Mask R-CNN, Retina Net, and Retina U-Net, tailored for medical image analysis. It includes training and inference pipelines with dynamic patching, prediction consolidation, and object/patient-level evaluation. The project is explicitly no longer maintained in favor of nnDetection.

## Use cases
- detect lesions in CT scans with 3D object detection
- train Mask R-CNN on medical images with bounding box annotations
- run instance segmentation on 2D radiology images
- evaluate object detection models on patient-level metrics
- apply Retina U-Net combining segmentation and detection supervision
- patch and tile large 3D medical volumes for training and inference

## When to choose
- you need 2D or 3D object detection on medical imaging data with a proven research framework
- you want to reproduce results from the Retina U-Net publication
- you need combined bounding box and pixel-wise annotation training

## When to avoid
- you need an actively maintained project - use nnDetection instead
- your use case is general-purpose object detection on natural images
- you require modern PyTorch versions or recent GPU support

## Facets
- artifact type: framework
- maturity: abandoned
- function: machine-learning, deep-learning, image-processing, computer-vision, data-science
- domain: deep-learning, computer-vision, healthcare, machine-learning
- platform: python
- tags: object-detection, medical-imaging, instance-segmentation, pytorch, 3d-detection, mask-rcnn, retina-unet, medical-image-analysis, linux, gpu

## Member repositories
- MIC-DKFZ/medicaldetectiontoolkit (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:29.525058+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:41:48.540402+00:00, confidence not recorded.
  - readme: https://github.com/MIC-DKFZ/medicaldetectiontoolkit (fetched 2026-08-28T04:04:29.525058+00:00, sha 4e05bb37710d)
- Data as of 2026-08-30T08:39:29.467469+00:00.