Skip to content

Release Notes

0.1.17_2.13.0 [2026-06-09]

Enhancements:

  • Added segment-lung-nodules config option to segment lung and lung nodules using the -ta lung_nodules task (CT only); mutually exclusive with segment-mr and use-fast

Fixes:

  • Changed default value of compute-statistics from false to true
  • Changed default value of use-fast from false to true

Maintenance:

  • Updated setuptools constraint from ==78.1.1 to >=80.10.1
  • Updated pytest-cov upper bound from <6 to <7
  • Updated PYVER environment variable from 3.11.11 to 3.11
  • Added TOTALSEG_WEIGHTS_PATH environment variable pointing to /opt/totalsegmentator_weights
  • Sorted and reorganized environment block in manifest.json

Documentation:

  • Updated README.md to document the new segment-lung-nodules option and its incompatibility constraints

0.1.16_2.13.0 [2026-05-29]

Enhancements:

  • Added output-nifti, output-dicom-seg, and output-dicom-rtstruct config options to control which output formats are produced
  • Extended Input-File to accept DICOM zip archives (.zip of a DICOM series) in addition to NIfTI files (.nii / .nii.gz)
  • Added startup validation that rejects incompatible input/output combinations (e.g., DICOM output with a NIfTI input) and enforces that at least one output type is enabled

Fixes:

  • Fixed runtime crash caused by dataclasses backport package (a Python 3.6 shim pulled in transitively by rt-utils) shadowing the stdlib on Python 3.11; the package is now explicitly uninstalled in the Dockerfile
  • Fixed missing GUI/X11 library error at runtime by replacing opencv-python with opencv-python-headless in the Dockerfile

Maintenance:

  • Added highdicom>=0.22,<1.0, rt-utils>=1.2,<2.0, and pydicom>=2.4,<4.0 dependencies to support DICOM output formats
  • Updated protobuf from 4.25.8 to 5.29.6
  • Refactored parser.py to return a typed GearConfig dataclass instead of a bare tuple
  • Refactored run.py into focused helper functions (_build_cmd, _prepare_input, _post_process, _run_subprocess)
  • Added pytest hook to .pre-commit-config.yaml
  • Added tests/test_run.py and tests/test_utils.py with comprehensive unit tests for command construction, subprocess handling, DICOM zip extraction, and post-processing

Documentation:

  • Updated README.md to document the new Input-File format options, the new output config flags, DICOM output constraints, and revised output file descriptions

0.1.15_2.13.0 [2026-03-27]

Maintenance:

  • Updated TotalSegmentator from 2.10.0 to 2.13.0
  • Removed automated release notes CI configuration

0.1.14_2.10.0 [2026-02-19]

Fixes:

  • Fixed model weights being re-downloaded at runtime by storing them in a fixed path (/opt/totalsegmentator_weights) during the Docker build and copying them into the final production image via TOTALSEG_WEIGHTS_PATH

0.1.13_2.10.0 [2025-09-23]

Enhancements:

  • Added MR image segmentation support via segment-mr config option
  • Added include-incomplete config option to include incomplete statistics in output

Fixes:

  • Changed input name from CT to Input-File to support both CT and MR images

Maintenance:

  • Downgraded python from 3.12 to 3.11.11
  • Changed base Docker image from nvcr.io/nvidia/pytorch:25.05-py3 to pytorch/pytorch:2.6.0-cuda12.6-cudnn9-runtime
  • Downgraded pytorch from 2.8.0a0+5228986 to 2.6.0
  • Updated setuptools to 78.1.1 for security patches
  • Renamed package from fw-gear-ct-total-segmentator to fw-gear-total-segmentator
  • Migrated from poetry to uv for dependency management

Documentation:

  • Updated README to clarify support for both CT and MR image segmentation
  • Updated gear description to reflect support for both CT and MR images
  • Removed outdated use case section from README

0.1.12_2.10.0 [2025-07-16]

Enhancements:

  • Added force-split configuration option to process images in chunks for reduced memory consumption
  • Added environment variables for UID (31337) and GID (31337) to support non-root execution
  • Enabled multi-threading control for resampling and saving operations (set to 1 thread to prevent out-of-memory issues)

Fixes:

  • Fixed security vulnerabilities by removing jupyter_core, protobuf, and pillow packages from base image
  • Fixed test coverage threshold from 80% to 50% to reflect actual coverage
  • Improved output streaming to capture stderr for better error diagnostics
  • Added proper exception handling and process cleanup in main execution loop

Maintenance:

  • Upgraded TotalSegmentator from 2.0.5 to 2.10.0
  • Upgraded base image from nvcr.io/nvidia/pytorch:23.05-py3 to nvcr.io/nvidia/pytorch:25.05-py3
  • Upgraded python from ^3.8 to ^3.12
  • Migrated from pip to uv for dependency management
  • Migrated from poetry to uv build system with PEP 621 format
  • Implemented multi-stage Docker build with separate base, build, dev, and production stages
  • Added non-root user (flywheel) for production container execution
  • Upgraded pytest from ^6.1.2 to ^8.3.2
  • Upgraded pytest-cov from ^3.0.0 to ^5.0.0
  • Upgraded pytest-mock from ^3.8.2 to ^3.14.0
  • Added pinned dependencies: six>=1.16,<1.17, protobuf==4.25.8, jupyter-core==5.8.1
  • Migrated CI configuration from sse-qa-ci to qa-ci project with sse branch
  • Updated pre-commit hooks from black to ruff_format and added ruff linter
  • Increased Trivy scan timeout to 3h and test timeout to 6h
  • Updated NVIDIA driver capabilities to include video capability
  • Removed several development packages (libc6-dev, linux-libc-dev) to reduce attack surface
  • Updated .dockerignore to include requirements-dev.txt
  • Disabled CI cache for test jobs to improve reliability

Documentation:

  • Fixed trailing whitespace and formatting issues throughout README and CONTRIBUTING files
  • Converted Poetry documentation URL from inline to reference format

0.1.11_2.0.5 [2024-11-14]

Enhancements:

  • Added automatic zipping of individual rib segmentation files when use-multilabel is disabled (generates all_right_ribs.zip and all_left_ribs.zip for easier handling of 100+ output files)
  • Added statistics JSON output file handling when compute-statistics is enabled

Fixes:

  • Fixed missing compute_statistics return value in parser function signature

Maintenance:

  • Upgraded acvl_utils to 0.2 and typing_extensions to latest versions
  • Optimized Dockerfile by consolidating TotalSegmentator installation steps into fewer layers
  • Improved Dockerfile readability with inline comments explaining installation steps

0.1.10_2.0.5 [2024-03-29]

Maintenance:

  • Updated CI/CD configuration to use larger runner resources for Docker operations
  • Added cache clearing capability to GitLab CI pipeline
  • Applied code formatting improvements across Python modules

0.1.9_2.0.5 [2023-12-09]

Maintenance:

  • Updated Python interpreter path from /opt/conda/bin/python to /usr/bin/python
  • Updated TotalSegmentator executable path from /opt/conda/bin/TotalSegmentator to /usr/local/bin/TotalSegmentator

0.1.8_2.0.5 [2023-12-06]

Enhancements:

  • Added resolution-based prefix (low_res or high_res) to combined segmentation output filenames for better identification of the model used

Fixes:

  • Fixed unused subprocess call when moving combined segmentation output file

Maintenance:

  • Updated base Docker image from pytorch:22.08-py3 to pytorch:23.05-py3
  • Updated TotalSegmentator from 1.5.2 to 2.0.5
  • Updated SimpleITK from 2.0.2 to 2.3.0
  • Added libarchive-tools system package installation
  • Removed isort and pytest hooks from pre-commit configuration
  • Changed gear category from analysis to utility

0.1.4_1.5.2 [2023-03-28]

Fixes:

  • Added SimpleITK==2.0.2 dependency to resolve segmentation processing issues

Maintenance:

  • Enabled debug mode in CI configuration for improved troubleshooting

0.1.2_1.5.2 [2023-03-10]

Enhancements:

  • Added CT segmentation gear wrapping TotalSegmentator deep learning algorithm for segmenting 104 anatomical structures in CT images
  • Implemented configuration options for fast mode (--fast flag), multilabel output (--ml flag), and statistics computation (--statistics flag)
  • Added support for GPU-accelerated processing using NVIDIA CUDA environment
  • Included automatic download and installation of TotalSegmentator v1.5.2 pretrained weights during Docker build

Maintenance:

  • Added CI/CD pipeline configuration with GitLab CI integration
  • Configured pre-commit hooks for code quality checks including gearcheck, poetry_export, docker_build, markdownlint, yamllint, black, isort, and pytest
  • Implemented Docker containerization based on nvcr.io/nvidia/pytorch:22.08-py3 base image
  • Added .dockerignore file to optimize Docker build context and reduce image size
  • Set up dependency management using poetry with flywheel-gear-toolkit v0.6+
  • Configured pytest with 80% minimum coverage requirement

Documentation:

  • Added comprehensive README with usage instructions, workflow diagrams, input specifications, and configuration options
  • Added CONTRIBUTING.md with development setup instructions, dependency management guidelines, and release process documentation
  • Included MIT license and citation information for the underlying TotalSegmentator algorithm