AVS-Net/knee_fast_mri

---
license: afl-3.0
tags:
- medical
size_categories:
- 100M<n<1B
---

# Dataset for AVS-Net Pre-training

The dataset utilized in the pre-training of the AVS-Net: Attention-based Variable Splitting Network for P-MRI Acceleration model, developed by Y Zhang, J Li, Z Wang, J Duan, and J Li, incorporates data from five distinct protocol sequences. These are:

- (coronal_pd)Coronal Spin Density-weighted without Fat Suppression
- (coronal_pd_fs)Coronal Spin Density-weighted with Fat Suppression
- (sagittal_pd)Sagittal Spin Density-weighted
- (sagittal_t2)Sagittal T2-weighted with Fat Suppression
- (axial_t2)Axial T2-weighted with Fat Suppression

The dataset is structured on a slice-by-slice basis, with each slice containing 20 cases. Each case is comprised of two files: rawdata*.mat and espirit*.mat. The dataset's structure can be outlined as follows:

## Dataset architecture:

- name: /rds/projects/d/duanj-ai-in-medical-imaging/knee_fast_mri
- Protocol: [coronal_pd, coronal_pd_fs, sagittal_pd, sagittal_t2, axial_t2]

Approximately 40 slices per protocol, each slice containing 15 channels, with a height and width (HW) of (640, 368)

```
knee_nyu
- axial_t2 coronal_pd(X) coronal_pd_fs sagittal_pd sagittal_t2
| | | | |
- [1, 2, 3, 4, 5, 6, 7, 8, 9, 10], [11, 12, 13, 14, 15, 16, 17, 18, 19, 20] masks
| |
- [train] [val]
| |
- espirit*.mat(1-40), rawdata*.mat(1-40) *_masks.mat
```

In this structure, each protocol has approximately 40 slices, each consisting of 15 channels. The dimensions of the data are 640x368 (height x width). For each protocol, the slices are further divided into two groups: the training set ([train]) and the validation set ([val]). The training set includes the espirit*.mat and rawdata*.mat files for each slice, while the validation set contains *_masks.mat files.

## Dataset Usage

> For a standalone knee dataset download, use `git lfs`(<https://git-lfs.com/>) to download from the `huggingface` datasets(<https://huggingface.co/datasets/AVS-Net/knee_fast_mri>):

```bash
# Make sure you have git-lfs installed (https://git-lfs.com)
git lfs install
git clone -j8 [e-mail]:datasets/AVS-Net/knee_fast_mri
```

## Known Issues and Resolutions

- 1. Network Connection Issue

For enhanced network connection quality, it is recommended to employ the `ssh` protocol instead of `https`.

```bash
# Rather than utilizing `https://huggingface.co/datasets/AVS-Net/knee_fast_mri`
# Clone the repository using `[e-mail]:datasets/AVS-Net/knee_fast_mri`
# As an example:
git clone -j8 [e-mail]:datasets/AVS-Net/knee_fast_mri
```

- 2. Interruptions During Download

Certain error messages may appear during the download process due to interruptions. These errors can include:

```
error: ... : cannot add to the index - missing --add option?
batch response: Post ... : read: connection reset by peer
error: failed to fetch some objects from 'https://hf.co/datasets/AVS-Net/knee_fast_mri.git/info/lfs'
```

Following the instructions below allows for the handling of these interruptions.

```bash
# Navigate (`cd`) to the directory containing the `lfs` folder
# Intead of using `git pull`,
# Use `git lfs pull` to resume the download progress for `lfs` projects

git lfs pull
```
Please note that this process will resume the download from where it was interrupted, thereby ensuring the integrity of your downloaded data.

原始信息汇总

数据集概述

数据集名称

Dataset for AVS-Net Pre-training

数据集描述

该数据集用于预训练AVS-Net模型，即Attention-based Variable Splitting Network for P-MRI Acceleration模型。数据集包含五个不同的协议序列：

• (coronal_pd) Coronal Spin Density-weighted without Fat Suppression
• (coronal_pd_fs) Coronal Spin Density-weighted with Fat Suppression
• (sagittal_pd) Sagittal Spin Density-weighted
• (sagittal_t2) Sagittal T2-weighted with Fat Suppression
• (axial_t2) Axial T2-weighted with Fat Suppression

数据集结构

数据集按切片组织，每片包含20个病例。每个病例包含两个文件：rawdata*.mat 和 espirit*.mat。数据集的结构如下：

数据集架构

• 名称: /rds/projects/d/duanj-ai-in-medical-imaging/knee_fast_mri
• 协议: [coronal_pd, coronal_pd_fs, sagittal_pd, sagittal_t2, axial_t2]

每个协议大约有40个切片，每个切片包含15个通道，高度和宽度（HW）为（640, 368）。

knee_nyu - axial_t2 coronal_pd(X) coronal_pd_fs sagittal_pd sagittal_t2 | | | | | - [1, 2, 3, 4, 5, 6, 7, 8, 9, 10], [11, 12, 13, 14, 15, 16, 17, 18, 19, 20] masks | | - [train] [val] | | - espirit*.mat(1-40), rawdata*.mat(1-40) *_masks.mat

数据集使用

数据集可以通过git lfs下载，具体命令如下：

bash

确保已安装git-lfs

git lfs install git clone -j8 [e-mail]:datasets/AVS-Net/knee_fast_mri

已知问题及解决方案

• 网络连接问题：建议使用ssh协议代替https。
• 下载中断问题：使用git lfs pull命令恢复下载。