WCODE-PIA: Partial-instance Annotation

WCODE-PIA, which focuses on the learning of incomplete annotations, is a medical image segmentation framework improved from WCODE-v0.

• This project focuses on the incomplete labeling task, in which the foreground area is partially labeled, and the remaining pixels are considered as the background.

---

📖 Our works

Title	Implementation	Web
Weakly Supervised Lymph Nodes Segmentation Based on Partial Instance Annotations with Pre-trained Dual-branch Network and Pseudo Label Learning	DBDMP	MELBA2024
ReCo-I2P: An Incomplete Supervised Lymph Node Segmentation Framework Based on Orthogonal Partial-Instance Annotation	ReCo-I2P	MICCAI2025 (Oral)
Learning from 3D Partial Foreground Annotations: Prototype-Enhanced Incomplete Supervision for Lymph Node Segmentation	ReCo-I2P+	-

---

🔬 Related Literatures

Some implementations of compared state-of-the-art (SOTA) methods can be found here.

IA - Inaccurate label, IC - Incomplete label

Category	Authors	Title	Implementation	Web
IA	B. Han et al.	Co-teaching: robust training of deep neural networks with extremely noisy labels	Coteaching	NeurIPS2018
IA	C. Fang et al.	Reliable Mutual Distillation for Medical Image Segmentation Under Imperfect Annotations	RMD	TMI2023
IA	T. Weng et al.	Accurate Segmentation of Optic Disc and Cup from Multiple Pseudo-labels by Noise-aware Learning	MPNN	CSCWD2024
IC	C. Liu et al.	AIO2: Online Correction of Object Labels for Deep Learning With Incomplete Annotation in Remote Sensing Image Segmentation	AIO2	TGRS2024
IC	H. Zhou et al.	Unsupervised domain adaptation for histopathology image segmentation with incomplete labels	SASN_IL	CBM2024

---

💾 Dataset and Related Weight of Models

We only provide the preprocessed dataset used in the experiment; see this file for details.

Related weight of models

Dataset	DSC (%)	ASSD (mm)	I-F1 (%)
LNQ2023 ($P_1$)	57.97±15.90	10.88±11.40	31.66±14.83

LNQ2023 - BaiduNetdisk: https://pan.baidu.com/s/1vyDE5N51vtCqLFIXH-Srcw?pwd=0319

📚 Citation

@article{wang2024weakly,
  title={Weakly Supervised Lymph Nodes Segmentation Based on Partial Instance Annotations with Pre-trained Dual-branch Network and Pseudo Label Learning},
  author={Wang, Litingyu and Qu, Yijie and Luo, Xiangde and Liao, Wenjun and Zhang, Shichuan and Wang, Guotai},
  journal={Machine Learning for Biomedical Imaging},
  volume={2},
  note={MICCAI 2023 LNQ challenge special issue},
  pages={1030--1047},
  year={2024}
}

@inproceedings{wang2025reco,
  title={ReCo-I2P: An Incomplete Supervised Lymph Node Segmentation Framework Based on Orthogonal Partial-Instance Annotation},
  author={Wang, Litingyu and Ye, Ping and Liao, Wenjun and Zhang, Shichuan and Zhang, Shaoting and Wang, Guotai},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={507--517},
  year={2025},
  organization={Springer}
}

✉️ Contact

--- Email: litingyuwang@gmail.com