# nnFilterMatch

**Repository Path**: max-liulin/nnFilterMatch

## Basic Information

- **Project Name**: nnFilterMatch
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-11-27
- **Last Updated**: 2025-11-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

Please read the guide for usage


# 🗂️ Main

The **Main** folder contains the core implementation of **nnFilterMatch** in two forms:

- **📓 Jupyter Notebook version**  
  Located in `notebooks/nnFilterMatch_demo.ipynb`.  
  This is a detail version for quickly trying out nnFilterMatch.  

- **🐍 Python package version**  
  Located in `Main/nnFilterMatch/`.  
  This includes the complete implementation of the nnFilterMatch framework as a training class (`nnUNetTrainer_nnfiltermatch.py`) that integrates directly with [nnU-Net v2](https://github.com/MIC-DKFZ/nnUNet).
  

## 📂 Benchmark Datasets

Other folders in this repository contain **benchmark dataset information** used in our experiments. 
These include dataset splits and configuration details. 
- **ACDC** – Automated Cardiac Diagnosis Challenge  
- **LA** – Left Atrium segmentation dataset  
- **SegTHOR** – Thoracic organ segmentation
- **Prostate** – Prostate segmentation


## 📊 Results on ACDC Dataset-Ablation

### 🧪 5% Labeled Data, lr: 5e-3, weight_deacy: 3e-3

| Method | Dice (%) | IoU (%) | 95HD (voxel) | ASD (voxel) |
|--------|----------|---------|-----------|----------|
| SL     | 82.96    | 73.01   | 9.52      | 2.69     |
| SSL    | 89.51    | 81.59   | 2.43      | 0.71     |
| **SSL_AL (nnFilterMatch)** | **89.56** | **81.63** | **2.39** | **0.63** |

---

### 🧪 10% Labeled Data, lr: 1e-2, weight_deacy: 3e-5

| Method | Dice (%) | IoU (%) | 95HD (voxel) | ASD (voxel) |
|--------|----------|---------|-----------|----------|
| SL     | 86.31    | 77.55   | 4.42      | 1.14     |
| SSL    | 89.58    | 81.65   | 2.41      | 0.69     |
| **SSL_AL (nnFilterMatch)** | **90.06** | **82.41** | **1.46** | **0.44** |



## 📚 Citation
1. B. Zhao, C. Wang, and S. Ding, “CrossMatch: Enhance Semi-Supervised Medical Image Segmentation with Perturbation Strategies and Knowledge Distillation,” IEEE Journal of Biomedical and Health Informatics, pp. 1–13, Jan. 2024, doi: 10.1109/jbhi.2024.3463711.
2. Y. Wang, Z. Li, L. Qi, Q. Yu, Y. Shi, and Y. Gao, “Balancing Multi-Target Semi-Supervised Medical Image Segmentation with Collaborative Generalist and Specialists,” IEEE Transactions on Medical Imaging, p. 1, Jan. 2025, doi: 10.1109/tmi.2025.3557537.
3.  X. Liu, W. Li, and Y. Yuan, “DifFRECT: Latent diffusion label rectification for semi-supervised medical image segmentation,” in Lecture notes in computer science, 2024, pp. 56–66. doi: 10.1007/978-3-031-72390-2_6.


## 🙏 Acknowledgments

This project makes use of several **publicly available datasets**.  
We gratefully acknowledge the data providers and research communities for making these resources accessible and supporting open science.  

Our implementation is built upon the **[nnU-Net](https://github.com/MIC-DKFZ/nnUNet)** framework  
(Isensee et al., *Nature Methods*, 2021). We sincerely thank the authors for releasing their code to the community,  
which has been instrumental in the development of this work.