PARASIDE: PARAnasal Segmentation for Imaging-based Disease Evaluation

This is the official repository of the paper "PARASIDE: An Automatic Paranasal Sinus Segmentation and Structure Analysis Tool for Magnetic Resonance Imaging"

PARASIDE enables the automated whole-paranasal sinus segmentation for T1-weighted MRI, extracting quantitative features.

Citation

If you are using PARASIDE, please cite the following:

@article{moller2025paraside,
  title={PARASIDE: An Automatic Paranasal Sinus Segmentation and Structure Analysis Tool for MRI},
  author={M{\"o}ller, Hendrik and Krautschick, Lukas and Atad, Matan and Graf, Robert and Busch, Chia-Jung and Beule, Achim and Scharf, Christian and Kaderali, Lars and Menze, Bjoern and Rueckert, Daniel and Paperlein, Fabian},
  journal={arXiv preprint arXiv:2501.14514},
  year={2025}
}

Warning

This is currently the pre-print and will be replaced by the peer-reviewed paper as soon as available!

Installation

The order of the following instructions is important!

1. Use Conda or Pip to create a venv for python 3.11, we are using conda for this example:

conda create --name paraside python=3.11
conda activate paraside
conda install pip

2. Go to https://pytorch.org/get-started/locally/ and install a correct pytorch version for your machine in your venv
3. Confirm that your pytorch package is working! Try calling these commands:

nvidia-smi

This should show your GPU and it's usage.

python -c "import torch; print(torch.cuda.is_available())"

This should throw no errors and return True

In order to use this code, you need to install the TPTBox (https://github.com/Hendrik-code/TPTBox) among other packages. See pyproject.toml for the required packages.

pip install TPTBox

You can also just install all requirements by calling while on the directory level of the pyproject.toml

pip install -e .

(Optionally) Download the latest model weights from the release section (https://github.com/Hendrik-code/paraside/releases).

Installation complete.

Using PARASIDE

Command-line-interface (CLI):

paraside --i <inputfile/folder> --m <model-version/path> [--cpu]

Or here an example code to run PARASIDE:

from TPTBox import NII
from src.load_model import load_model
from src.inference_function import segment_nose
from src.ethmoid_split import split_ethmoid
from src.calc_features import compute_features

# load the image you want to process
input_file = "<path-to-your-file>.nii.gz"
input_nii: NII = NII.load(input_file, seg=False)

# load the segmentation model
paraside_segmentation_model = load_model(model_path_or_version)

# create the segmentation mask
segmentation: NII = segment_nose(
    image_nii=input_nii,
    model=paraside_segmentation_model,
)
# save the nifti back to disk
segmentation.save(output_dir + f"{input_filename}_segmentation.nii.gz")

# split the ethmoid segmentation into anterior and posterior part
segmentation_ethmoid_split = split_ethmoid(segmentation)
# save the nifti back to disk
segmentation_ethmoid_split.save(output_dir + f"{input_filename}_segmentation_ethmoid_split.nii.gz")

# take measurements
features: dict[str, float] = compute_features(
    input_nii,
    segmentation_ethmoid_split,
)

Tip

You can run the feature computation on masks where the ethmoid has not been splitted. It will then treat the whole ethmoid as one structure instead of three. Don't mix these, because that is not comparable with most features.

Segmentation Label

Label	Structure
1	Air Maxillaris right
2	Air Maxillaris left
3	Air Frontalis right
4	Air Frontalis left
5	Air Sphenoid right
6	Air Sphenoid left
7	Air Ethmoid right
8	Air Ethmoid left

If the ethmoid got split into multiple subparts, the labels are:

Label	Structure
7	Air Ethmoid Anterior right
8	Air Ethmoid Anterior left
107	Air Ethmoid Posterior right
108	Air Ethmoid Posterior left
207	Air Osteometeal Complex right
208	Air Osteometeal Complex left

The labels above are all the AIR labels. The corresponding SOFT TISSUE labels can be gathered by adding 8 on the airlabel.

e.G.:

Label	Structure
9	Soft tissue Maxillaris right
10	Soft tissue Maxillaris left
11	Soft tissue Frontalis right
12	Soft tissue Frontalis left
13	Soft tissue Sphenoid right
14	Soft tissue Sphenoid left
15	Soft tissue Ethmoid right
16	Soft tissue Ethmoid left

Contributing

If you want to contribute, create an issue or merge request. Best to start with creating an issue and discussing the proposed changes with us before implementation.

Authorship

This pipeline was created in a collaboration by:

Hendrik Möller (he/him)	Dr. Fabian Paperlein (he/him)
PhD Researcher	Medical Doctor
Department for Interventional and Diagnostic Neuroradiology	Department of Otorhinolaryngology, Head and Neck Surgery
University Hospital rechts der Isar at Technical University of Munich Ismaninger Street 22, 81675 Munich	University Medicine Greifswald Fleischmann Straße 10, Greifswald, 17475, Mecklenburg-Western
- https://deep-spine.de/ - https://aim-lab.io/author/hendrik-moller/	- https://www.researchgate.net/profile/Fabian-Paperlein - https://www.medizin.uni-greifswald.de/hno/klinik/mitarbeitende/schwitzing-fabian/ - https://www.medizin.uni-greifswald.de/hno/forschung/ki-basierte-diagnostik/
hendrik.moeller[at]tum.de	fabian.paperlein[at]med.uni-greifswald.de

We thank all co-authors and contributors!

License

Copyright 2025 Hendrik Möller

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.