# RP3D-DiagDS

**Overview of RP3D-DiagDS.** There are **39,026 cases (192,675 scans)** across 7 human anatomy regions and 9 diverse modalities covering **930 ICD-10-CM codes**.

<img src="https://github.com/qiaoyu-zheng/RP3D-Diag/blob/main/Images/RP3D-DiagDS.png"/>

The images used in our dataset can be downloaded from [BaiduYun](https://pan.baidu.com/s/1E_uSoCLm5H66a7KkpRfi1g?pwd=urfg)

## About Dataset

There are totally 4 json files:

1. RP3D_train.json: Data used for model training. This file is organized at case level (there may be more than one kind of modality and anatomy in a case. For more details, refer to the paper [Large-scale Long-tailed Disease Diagnosis on Radiology](https://qiaoyu-zheng.github.io/RP3D-Diag)).
2. RP3D_test_json: Data used for model evaluation.
3. disorder_label_dict.json: For disorder granularity. There are totally 5569 ( 5568 abnormal and 1 noraml) label. There disorders are sorted in descending order based on the corresponding case number for evaluation.
4. icd10_label_dict.json. For ICD-10-CM granularity. There are totally 931 ( 930 abnormal and 1 noraml) label. There disorders are sorted in descending order based on the corresponding case number for evaluation.

## About Checkpoint

The Checkpoint is in the file `checkpoint_32_late`, The detailed parameter we use for training is in the following:

```
start_class: 0
end_clas: 5569
backbone: 'resnet'
level: 'articles'
depth: 32
ltype: 'MultiLabel'
augment: True
split: 'late'
```

Please refer to our instructions on [github](https://github.com/qiaoyu-zheng/RP3D-Diag) to download and use.