README.md

---

license: cc-by-4.0
task_categories:
- image-classification
language:
- en
tags:
- x-ray
- medical
- chest
size_categories:
- 100K<n<1M
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: val
    path: data/val-*
- config_name: original
  data_files:
  - split: train
    path: original/train-*
  - split: test
    path: original/test-*
  - split: val
    path: original/val-*
dataset_info:
- config_name: default
  features:
  - name: UID
    dtype: string
  - name: Fold
    dtype: int64
  - name: Split
    dtype: string
  - name: PatientID
    dtype: string
  - name: PhysicianID
    dtype: string
  - name: StudyDate
    dtype: string
  - name: Age
    dtype: int64
  - name: Sex
    dtype: string
  - name: HeartSize
    dtype: int64
  - name: PulmonaryCongestion
    dtype: int64
  - name: PleuralEffusion_Right
    dtype: int64
  - name: PleuralEffusion_Left
    dtype: int64
  - name: PulmonaryOpacities_Right
    dtype: int64
  - name: PulmonaryOpacities_Left
    dtype: int64
  - name: Atelectasis_Right
    dtype: int64
  - name: Atelectasis_Left
    dtype: int64
  - name: Image
    dtype: image
  splits:
  - name: train
    num_bytes: 36724515176.076
    num_examples: 137593
  - name: test
    num_bytes: 11088307165.008
    num_examples: 42928
  - name: val
    num_bytes: 9210192401.0
    num_examples: 34860
  download_size: 58343808539
  dataset_size: 57023014742.084
- config_name: original
  features:
  - name: UID
    dtype: string
  - name: Fold
    dtype: int64
  - name: Split
    dtype: string
  - name: PatientID
    dtype: string
  - name: PhysicianID
    dtype: string
  - name: StudyDate
    dtype: string
  - name: Age
    dtype: int64
  - name: Sex
    dtype: string
  - name: HeartSize
    dtype: int64
  - name: PulmonaryCongestion
    dtype: int64
  - name: PleuralEffusion_Right
    dtype: int64
  - name: PleuralEffusion_Left
    dtype: int64
  - name: PulmonaryOpacities_Right
    dtype: int64
  - name: PulmonaryOpacities_Left
    dtype: int64
  - name: Atelectasis_Right
    dtype: int64
  - name: Atelectasis_Left
    dtype: int64
  - name: Image
    dtype: image
  splits:
  - name: train
    num_bytes: 793575463284.632
    num_examples: 137593
  - name: test
    num_bytes: 235100370576.352
    num_examples: 42928
  - name: val
    num_bytes: 197760028732.64
    num_examples: 34860
  download_size: 1266898242525
  dataset_size: 1226435862593.624
---


# TAIX-Ray Dataset

TAIX-Ray is a comprehensive dataset of approximately 200k bedside chest radiographs from around 50k intensive care patients at University Hospital Aachen, Germany, collected between 2010 and 2024.

Trained radiologists provided structured reports at the time of acquisition, assessing key findings such as cardiomegaly, pulmonary congestion, pleural effusion, pulmonary opacities, and atelectasis on an ordinal scale.

<br>

## Code & Details

The code for data loading, preprocessing, and baseline experiments is available at:
[https://github.com/TruhnLab/TAIX-Ray.git](https://github.com/TruhnLab/TAIX-Ray.git)

---

## How to Use

### Prerequisites

Ensure you have the following dependencies installed:

```bash
pip install datasets matplotlib huggingface_hub pandas tqdm
```

---

## Configurations

This dataset is available in two configurations:

| **Name** | **Size** | **Image Size** |
| -------- | -------- | -------------- |
| default  | 62GB     | 512px          |
| original | 1.2TB    | variable       |

---

## Option A: Use within the Hugging Face Framework

If you want to use the dataset directly within the Hugging Face `datasets` library, you can load and visualize it as follows:

```python
from datasets import load_dataset
from matplotlib import pyplot as plt

# Load the TAIX-Ray dataset
dataset = load_dataset("TLAIM/TAIX-Ray", name="default")

# Access the training split (Fold 0)
ds_train = dataset["train"]

# Retrieve a single sample from the training set
item = ds_train[0]

# Extract and display the image
image = item["Image"]
plt.imshow(image, cmap="gray")
plt.savefig("image.png")  # Save the image to a file
plt.show()  # Display the image

# Print metadata (excluding the image itself)
for key in item.keys():
    if key != "Image":
        print(f"{key}: {item[key]}")
```

---

## Option B: Downloading the Dataset

If you prefer to download the dataset to a specific folder, use the following script. This will create the following folder structure:

```
.
├── data/
│   ├── 549a816ae020fb7da68a31d7d62d73c418a069c77294fc084dd9f7bd717becb9.png
│   ├── d8546c6108aad271211da996eb7e9eeabaf44d39cf0226a4301c3cbe12d84151.png
│   └── ...
└── metadata/
    ├── annotation.csv
    └── split.csv
```

```python
from datasets import load_dataset
from pathlib import Path
import pandas as pd
from tqdm import tqdm

# Define output paths
output_root = Path("./TAIX-Ray")

# Create folders
data_dir = output_root / "data"
metadata_dir = output_root / "metadata"
data_dir.mkdir(parents=True, exist_ok=True)
metadata_dir.mkdir(parents=True, exist_ok=True)

# Load dataset in streaming mode
dataset = load_dataset("TLAIM/TAIX-Ray", name="default", streaming=True)

# Process dataset
metadata = []
for split, split_dataset in dataset.items():
    print("-------- Start Download:", split, "--------")
    for item in tqdm(split_dataset, desc="Downloading"):  # Stream data one-by-one
        uid = item["UID"]
        img = item.pop("Image")  # PIL Image object

        # Save image
        img.save(data_dir / f"{uid}.png", format="PNG")

        # Store metadata
        metadata.append(item)

# Convert metadata to DataFrame
metadata_df = pd.DataFrame(metadata)

# Save annotations to CSV file
metadata_df.drop(columns=["Split", "Fold"]).to_csv(
    metadata_dir / "annotation.csv", index=False
)

print("Dataset streamed and saved successfully!")
```