This repository contains code to analyse the output of TokenUNet training. TokenUNet is trained with nnunetv2. Insert the custom folder into /your/path/nnUNet/nnunetv2/training/nnUNetTrainer/
Research use only. Not a medical device. This software was developed in the HEREDITARY project as a research prototype. It has not been assessed or certified under the EU Medical Device Regulation (EU) 2017/745. It must not be used to diagnose, treat, monitor or make any other decision about individual patients.
| Field | Value |
|---|---|
| Name | TokenUNet |
| Version | 0.1 |
| Type | Software or library |
| Owner partner and contact | University of Padua, louisfabrice.tshimanga@unipd.it |
| License | SPDX identifier (see LICENSE file) |
| Repository and DOI | https://doi.org/10.1371/journal.pone.0354511 |
| Release date | 17.08.2026 |
| Related task or deliverable | T4.3 |
| How to cite | https://doi.org/10.1371/journal.pone.0354511 |
Primary intended use. This repository contains code to analyse the output of TokenUNet training.
Intended users. Biomedical researchers, deep learning scientists.
Out-of-scope uses. This release must not be used for
- clinical diagnosis, prognosis or treatment of individual patients;
- decisions about individuals, such as insurance, employment or access to care;
- re-identification of individuals or linkage with other data sources;
- populations, devices or data types not described in Section 3.
| Dataset | Source partner | Type (public, synthetic, pseudonymized, anonymized) | Population (size, age, sex, country, condition) | Ethics approval or access conditions |
|---|---|---|---|---|
| FeTS 2022 | NA | Public, anonymized | 1251 glioma patients | see https://www.synapse.org/Synapse:syn28546456/wiki/617093 |
This repository and the released files contain no personal data, pseudonymized records, identifiers or pseudonymization keys.
- Algorithm or architecture. Convolutional Neural Network, Transformer
- Training setting. Centralized
- Privacy techniques. none
- Evaluation data. FeTS Cross-validation.
- Metrics and results. 0.81-0.85 Mean Dice.
- Results by subgroup. NA
- Generalizability. NA
- Known biases and confounders. Similarity to training data.
- Failure modes. NA
- Privacy risks. NA
- Terms of use. research use-only.
- Contact for misuse or vulnerability reports. louisfabrice.tshimanga@unipd.it, manfredo.atzori@unipd.it
Funded by the European Union under grant agreement No 101137074 (HEREDITARY). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Health and Digital Executive Agency (HaDEA). Neither the European Union nor the granting authority can be held responsible for them.