{"architecture":{"family":"U-Net (2D, 3D full-res, 3D cascade), auto-configured","input":"NIfTI volumes + label maps","output":"label maps","pretraining":"none \u2014 trained from scratch per dataset"},"article":null,"cancer_slugs":["pan-cancer"],"category":"radiology","confidence":"high","datasets":[{"name":"BraTS \u2014 Brain Tumor Segmentation Challenge","note":"standard glioma segmentation benchmark where nnU-Net variants placed first repeatedly","role":"benchmark","slug":"brats"}],"developer":"German Cancer Research Center (DKFZ), Division of Medical Image Computing","evaluation":[{"benchmark":"Medical Segmentation Decathlon + 23 public challenges (paper)","external":true,"metric":"Dice","source":"https://www.nature.com/articles/s41592-020-01008-z","value":"first place or top tier on the majority"}],"hf":null,"kind":"framework","license":"apache-2.0","limitations":"- Needs voxel-level labels; no pretraining shortcut.\n- Long training times; the cascade variant is heavy.","links":{"demo":null,"docs":null,"doi":"10.1038/s41592-020-01008-z","github":"https://github.com/MIC-DKFZ/nnUNet","huggingface":null,"paper":"https://www.nature.com/articles/s41592-020-01008-z","pmid":null},"modalities":["radiology-ct","radiology-mri","pet"],"name":"nnU-Net","notable_uses":"","openness":"open-weights","regulatory":{"intended_use_en":"Framework for research; products built on it need their own clearance.","intended_use_pl":"Framework badawczy; produkty zbudowane na nim wymagaj\u0105 w\u0142asnego dopuszczenia.","source_url":"https://github.com/MIC-DKFZ/nnUNet","status":"not-applicable"},"regulatory_status":"not-applicable","release_date":"2020-12-07","run_snippet":"pip install nnunetv2\nnnUNetv2_plan_and_preprocess -d 001 --verify_dataset_integrity\nnnUNetv2_train 001 3d_fullres 0\nnnUNetv2_predict -i imagesTs/ -o preds/ -d 001 -c 3d_fullres","settings":["treatment-planning","basic-research"],"slug":"nnunet","sources":[{"label":"Isensee F et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 2021","url":"https://www.nature.com/articles/s41592-020-01008-z"},{"label":"Isensee F et al. nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation (preprint)","url":"https://arxiv.org/abs/1809.10486"},{"label":"GitHub \u2014 MIC-DKFZ/nnUNet","url":"https://github.com/MIC-DKFZ/nnUNet"}],"summary":"Self-configuring segmentation framework: given a labelled dataset it chooses preprocessing, network topology and training schedule automatically, and it remains the baseline to beat on most medical segmentation challenges, including tumour tasks.","tasks":["segmentation"],"training":{"summary_en":"No fixed training corpus. nnU-Net ships no pretrained weights: it configures itself and trains from scratch on whatever labelled dataset the user supplies. In the original work it was applied without any manual adjustment between datasets to the Medical Segmentation Decathlon (10 tasks) and to further biomedical segmentation challenges.","summary_pl":"Brak sta\u0142ego korpusu treningowego. nnU-Net nie udost\u0119pnia wag wytrenowanych wcze\u015bniej: konfiguruje si\u0119 sam i trenuje od zera na zbiorze etykietowanym dostarczonym przez u\u017cytkownika. W pracy \u017ar\u00f3d\u0142owej zastosowano go bez \u017cadnych r\u0119cznych korekt mi\u0119dzy zbiorami do Medical Segmentation Decathlon (10 zada\u0144) oraz kolejnych konkurs\u00f3w segmentacji biomedycznej."},"updated_at":"2026-09-06T15:20:45.307197","url":"/ai-oncology/models/nnunet","usage":{"hardware":"GPU with \u226510 GB for 3D full-resolution training","library":"pip package nnunetv2","snippet":"pip install nnunetv2\nnnUNetv2_plan_and_preprocess -d 001 --verify_dataset_integrity\nnnUNetv2_train 001 3d_fullres 0\nnnUNetv2_predict -i imagesTs/ -o preds/ -d 001 -c 3d_fullres"},"verified_at":"2026-09-05T22:26:01.557400","verified_by":"editorial","version":"v2","what_it_does":"Train a strong 2D/3D U-Net on your own annotated CT/MRI/PET volumes without architecture engineering; used for tumour, organ-at-risk and lymph-node segmentation in radiotherapy and radiomics research."}