TorchXRayVision

https://mlmed.org/torchxrayvision/
Admin
TorchXRayVision is an open source Python library providing pre-trained deep learning models and...
Tokens:12,858
Snippets:98
Trust Score:7.9
Update:6 months ago
Show doc for...
Tokens:
Raw
### Install TorchXRayVision

Source: https://mlmed.org/torchxrayvision/index.html

Command to install the package via pip.

```bash
pip install torchxrayvision
```

--------------------------------

### Process Image and Run Inference

Source: https://mlmed.org/torchxrayvision/index.html

Example demonstrating image normalization, transformation, and model inference using a pre-trained DenseNet model.

```python
import torchxrayvision as xrv
import skimage, torch, torchvision

# Prepare the image:
img = skimage.io.imread("16747_3_1.jpg")
img = xrv.datasets.normalize(img, 255) # convert 8-bit image to [-1024, 1024] range
img = img.mean(2)[None, ...] # Make single color channel

transform = torchvision.transforms.Compose([
    xrv.datasets.XRayCenterCrop(),
    xrv.datasets.XRayResizer(224),
])

img = transform(img)
img = torch.from_numpy(img)

# Load model and process image
model = xrv.models.DenseNet(weights="densenet121-res224-all")
outputs = model(img[None,...]) # or model.features(img[None,...])

# Print results
dict(zip(model.pathologies, outputs[0].detach().numpy()))
```

--------------------------------

### Display Dataset Representation

Source: https://mlmed.org/torchxrayvision/datasets.html

Example output of the __repr__ method for a dataset, showing sample counts and views.

```text
CheX_Dataset num_samples=191010 views=['PA', 'AP']
```

--------------------------------

### Display Pathology Counts in Jupyter

Source: https://mlmed.org/torchxrayvision/datasets.html

Example output of pathology counts when the dataset is inspected within a Jupyter notebook environment.

```python
{'Atelectasis': {0.0: 17621, 1.0: 29718},
 'Cardiomegaly': {0.0: 22645, 1.0: 23384},
 'Consolidation': {0.0: 30463, 1.0: 12982},
 ...}
```

--------------------------------

### Load VinBrain Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Instantiate the VinBrain dataset by providing the image path and CSV path. Ensure the paths point to the correct dataset files.

```python
d_vin = xrv.datasets.VinBrain_Dataset(
    imgpath=".../train",
    csvpath=".../train.csv"
)

```

--------------------------------

### Class: VinBrain_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Initializes the VinBrain Dataset loader.

```APIDOC
## VinBrain_Dataset

### Description
Loads the VinDr-CXR dataset of chest X-rays with radiologist annotations.

### Parameters
- **imgpath** (str) - Required - Path to the image directory.
- **csvpath** (str) - Optional - Path to the CSV file (default: 'USE_INCLUDED_FILE').
- **views** (list, optional) - Optional - List of views to include.
- **transform** (callable, optional) - Optional - Transformation function.
- **data_aug** (callable, optional) - Optional - Data augmentation function.
- **seed** (int) - Optional - Random seed (default: 0).
- **pathology_masks** (bool) - Optional - Whether to include pathology masks (default: False).
```

--------------------------------

### Initialize 224x224 DenseNet Models

Source: https://mlmed.org/torchxrayvision/models.html

Instantiate DenseNet models with various pre-trained weight configurations for 224x224 input images.

```python
model = xrv.models.DenseNet(weights="densenet121-res224-all")
model = xrv.models.DenseNet(weights="densenet121-res224-rsna") # RSNA Pneumonia Challenge
model = xrv.models.DenseNet(weights="densenet121-res224-nih") # NIH chest X-ray8
model = xrv.models.DenseNet(weights="densenet121-res224-pc") # PadChest (University of Alicante)
model = xrv.models.DenseNet(weights="densenet121-res224-chex") # CheXpert (Stanford)
model = xrv.models.DenseNet(weights="densenet121-res224-mimic_nb") # MIMIC-CXR (MIT)
model = xrv.models.DenseNet(weights="densenet121-res224-mimic_ch") # MIMIC-CXR (MIT)
```

--------------------------------

### Initialize 512x512 ResNet Models

Source: https://mlmed.org/torchxrayvision/models.html

Instantiate a ResNet model configured for 512x512 input images.

```python
# 512x512 models
model = xrv.models.ResNet(weights="resnet50-res512-all")
```

--------------------------------

### Class: ObjectCXR_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Initializes the ObjectCXR Dataset loader.

```APIDOC
## ObjectCXR_Dataset

### Description
Loads the ObjectCXR dataset containing chest X-rays with annotations of foreign objects.

### Parameters
- **imgzippath** (str) - Required - Path to the image zip file.
- **csvpath** (str) - Required - Path to the CSV file.
- **transform** (callable, optional) - Optional - Transformation function.
- **data_aug** (callable, optional) - Optional - Data augmentation function.
- **seed** (int) - Optional - Random seed (default: 0).
```

--------------------------------

### Load and run Riken age estimation model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the AgeModel to estimate patient age from a chest X-ray. The model expects 320x320 input resolution.

```python
model = xrv.baseline_models.riken.AgeModel()

image = xrv.utils.load_image('00027426_000.png')
image = torch.from_numpy(image)[None,...]

pred = model(image)
# tensor([[50.4033]], grad_fn=<AddmmBackward0>)
```

--------------------------------

### Mira Sex Model Initialization and Usage

Source: https://mlmed.org/torchxrayvision/models.html

This snippet demonstrates how to initialize the Mira Sex Model, load an image, and obtain a sex prediction. The model automatically scales images to its native resolution of 224x224.

```APIDOC
## Mira Sex Model API

### Description
This API provides access to the Mira Sex Model, a ResNet34-based model trained to predict patient sex from chest X-rays using the CheXpert dataset. The model's native resolution is 224x224, and images are automatically scaled.

### Method
POST (Implicit, as model is called with image data)

### Endpoint
`/websites/mlmed_torchxrayvision/baseline_models/mira/SexModel`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None (Model is initialized and then called with image tensor)

### Initialization Example
```python
import torch
import torchxrayvision as xrv

# Initialize the model with pre-trained weights
model = xrv.baseline_models.mira.SexModel(weights=True)
```

### Prediction Example
```python
# Load and preprocess an image
image = xrv.utils.load_image('00027426_000.png')
image = torch.from_numpy(image)[None,...] # Add batch dimension

# Get prediction
pred = model(image)

# Get the predicted sex label
predicted_sex = model.targets[torch.argmax(pred)]

print(f"Predicted Sex: {predicted_sex}")
# Expected output: 'Male' or 'Female'
```

### Response
#### Success Response (200)
- **prediction** (Tensor) - A tensor representing the model's output probabilities for each sex category.
- **targets** (List[str]) - A list of possible sex labels, e.g., ['Male', 'Female'].

#### Response Example
```json
{
  "prediction": [0.1, 0.9], 
  "targets": ["Male", "Female"]
}
```

### Publication
Algorithmic encoding of protected characteristics in chest X-ray disease detection models B. Glocker, C. Jones, M. Bernhardt, S. Winzeck eBioMedicine. Volume 89, 104467, 2023.

### Related Project
Chexploration: Medical Image Representation and Analysis (MIRA Team, 2023) - https://github.com/biomedia-mira/chexploration
```

--------------------------------

### Load and run ChestX-Det segmentation model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the PSPNet segmentation model and processes an image to obtain anatomical masks.

```python
seg_model = xrv.baseline_models.chestx_det.PSPNet()
output = seg_model(image)
output.shape # [1, 14, 512, 512]
seg_model.targets # ['Left Clavicle', 'Right Clavicle', 'Left Scapula', 'Right Scapula',
                  #  'Left Lung', 'Right Lung', 'Left Hilus Pulmonis', 'Right Hilus Pulmonis',
                  #  'Heart', 'Aorta', 'Facies Diaphragmatica', 'Mediastinum',  'Weasand', 'Spine']
```

--------------------------------

### ResNet Models (512x512)

Source: https://mlmed.org/torchxrayvision/models.html

Instantiate ResNet models with pre-trained weights for 512x512 image resolution. These models are based on the 'Deep Residual Learning for Image Recognition' paper.

```APIDOC
## ResNet Models (512x512)

### Description
Instantiate ResNet models with pre-trained weights for 512x512 image resolution. These models are based on the 'Deep Residual Learning for Image Recognition' paper.

### Method
`xrv.models.ResNet(weights, cache_dir, op_threshs, apply_sigmoid)`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
import torchxrayvision as xrv

# Example for ResNet with specific weights
model = xrv.models.ResNet(weights="resnet50-res512-all")
```

### Parameters
* **weights** (str) - Specify a weight name to load pre-trained weights.
* **cache_dir** (str) - Override where the weights will be stored (default is ~/.torchxrayvision/).
* **op_threshs** (dict) - Dictionary of thresholds for each target.
* **apply_sigmoid** (bool) - Apply a sigmoid function to the output.

### Targets
['Atelectasis', 'Consolidation', 'Infiltration', 'Pneumothorax', 'Edema', 'Emphysema', 'Fibrosis', 'Effusion', 'Pneumonia', 'Pleural_Thickening', 'Cardiomegaly', 'Nodule', 'Mass', 'Hernia', 'Lung Lesion', 'Fracture', 'Lung Opacity', 'Enlarged Cardiomediastinum']
```

--------------------------------

### Class: SIIM_Pneumothorax_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Initializes the SIIM Pneumothorax Dataset loader.

```APIDOC
## SIIM_Pneumothorax_Dataset

### Description
Loads the SIIM Pneumothorax Dataset containing DICOM images and RLE masks.

### Parameters
- **imgpath** (str) - Required - Path to the image directory.
- **csvpath** (str) - Optional - Path to the CSV file (default: 'USE_INCLUDED_FILE').
- **transform** (callable, optional) - Optional - Transformation function.
- **data_aug** (callable, optional) - Optional - Data augmentation function.
- **seed** (int) - Optional - Random seed (default: 0).
- **unique_patients** (bool) - Optional - Whether to filter for unique patients (default: True).
- **pathology_masks** (bool) - Optional - Whether to include pathology masks (default: False).
```

--------------------------------

### Load and run Xinario view classification model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the ViewModel to classify the view of a chest X-ray image.

```python
model = xrv.baseline_models.xinario.ViewModel()

image = xrv.utils.load_image('00027426_000.png')
image = torch.from_numpy(image)[None,...]

pred = model(image)
# tensor([[17.3186, 26.7156]]), grad_fn=<AddmmBackward0>)

model.targets[pred.argmax()]
# Lateral
```

--------------------------------

### DenseNet Models (224x224)

Source: https://mlmed.org/torchxrayvision/models.html

Instantiate DenseNet models with various pre-trained weights for 224x224 image resolution. These models are trained on different datasets for specific tasks.

```APIDOC
## DenseNet Models (224x224)

### Description
Instantiate DenseNet models with various pre-trained weights for 224x224 image resolution. These models are trained on different datasets for specific tasks.

### Method
`xrv.models.DenseNet(weights, cache_dir, op_threshs, apply_sigmoid)`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
import torchxrayvision as xrv

# Example for DenseNet with different weights
model = xrv.models.DenseNet(weights="densenet121-res224-all")
model_rsna = xrv.models.DenseNet(weights="densenet121-res224-rsna") # RSNA Pneumonia Challenge
model_nih = xrv.models.DenseNet(weights="densenet121-res224-nih") # NIH chest X-ray8
model_pc = xrv.models.DenseNet(weights="densenet121-res224-pc") # PadChest (University of Alicante)
model_chex = xrv.models.DenseNet(weights="densenet121-res224-chex") # CheXpert (Stanford)
model_mimic_nb = xrv.models.DenseNet(weights="densenet121-res224-mimic_nb") # MIMIC-CXR (MIT)
model_mimic_ch = xrv.models.DenseNet(weights="densenet121-res224-mimic_ch") # MIMIC-CXR (MIT)
```

### Parameters
* **weights** (str) - Specify a weight name to load pre-trained weights.
* **cache_dir** (str) - Override where the weights will be stored (default is ~/.torchxrayvision/).
* **op_threshs** (dict) - Dictionary of thresholds for each target.
* **apply_sigmoid** (bool) - Apply a sigmoid function to the output.

### Targets
['Atelectasis', 'Consolidation', 'Infiltration', 'Pneumothorax', 'Edema', 'Emphysema', 'Fibrosis', 'Effusion', 'Pneumonia', 'Pleural_Thickening', 'Cardiomegaly', 'Nodule', 'Mass', 'Hernia', 'Lung Lesion', 'Fracture', 'Lung Opacity', 'Enlarged Cardiomediastinum']
```

--------------------------------

### XRV ResNet Autoencoder

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the XRV ResNet Autoencoder model.

```APIDOC
## XRV ResNet Autoencoder

### Class
`torchxrayvision.models.ResNetAE`

### Description
This class implements a ResNet-based autoencoder for X-ray images.
```

--------------------------------

### Class: NLMTB_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Initializes the National Library of Medicine Tuberculosis Dataset loader.

```APIDOC
## NLMTB_Dataset

### Description
Loads the National Library of Medicine Tuberculosis Datasets (Montgomery County and Shenzhen).

### Parameters
- **imgpath** (str) - Required - Path to the image directory.
- **transform** (callable, optional) - Optional - Transformation function.
- **data_aug** (callable, optional) - Optional - Data augmentation function.
- **seed** (int) - Optional - Random seed (default: 0).
- **views** (list) - Optional - List of views to include (default: ['PA']).
```

--------------------------------

### Xinario View Model

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the Xinario View model.

```APIDOC
## Xinario View Model

### Class
`torchxrayvision.models.ViewModel`

### Description
This model is used for predicting the view of an X-ray image, likely from Xinario data.

**Attributes:**
- `targets: List[str]`: A list of view predictions.
```

--------------------------------

### Riken Age Model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the AgeModel to estimate patient age from chest X-rays.

```APIDOC
## AgeModel

### Description
Predicts patient age from chest X-rays. Images are automatically scaled to 320x320.

### Response
- **targets** (list) - ['Age']
```

--------------------------------

### ChestX-Det Segmentation Model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the PSPNet model for anatomical segmentation of chest X-rays.

```APIDOC
## PSPNet Segmentation Model

### Description
Loads a pretrained anatomical segmentation model for chest X-rays.

### Parameters
#### Initialization Parameters
- **cache_dir** (str) - Optional - Override directory used to store cached weights (default: ~/.torchxrayvision/)

### Response
- **output** (tensor) - Shape [1, 14, 512, 512] representing segmentation masks.
- **targets** (list) - List of 14 anatomical structures: ['Left Clavicle', 'Right Clavicle', 'Left Scapula', 'Right Scapula', 'Left Lung', 'Right Lung', 'Left Hilus Pulmonis', 'Right Hilus Pulmonis', 'Heart', 'Aorta', 'Facies Diaphragmatica', 'Mediastinum', 'Weasand', 'Spine']
```

--------------------------------

### Dataset Class Methods

Source: https://mlmed.org/torchxrayvision/datasets.html

Common methods available across all dataset implementations in the library.

```APIDOC
## Method: totals()

### Description
Compute counts of pathologies present in the dataset.

### Response
- **Returns** (Dict[str, Dict[str, int]]) - A dictionary containing pathology name mapped to a dictionary of label values and their counts.

## Method: __repr__()

### Description
Returns the name and a description of the dataset, including number of samples and views. In Jupyter environments, it also prints pathology counts.
```

--------------------------------

### Riken Age Model

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the Riken Age model.

```APIDOC
## Riken Age Model

### Class
`torchxrayvision.models.AgeModel`

### Description
A model developed by Riken for age prediction from X-ray images.

**Attributes:**
- `targets: List[str]`: A list of age-related predictions.
```

--------------------------------

### ChestX-Det Segmentation

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the ChestX-Det segmentation model.

```APIDOC
## ChestX-Det Segmentation

### Class
`torchxrayvision.models.PSPNet`

### Description
Implements the PSPNet architecture for segmentation tasks, specifically for ChestX-Det.

**Attributes:**
- `targets: List[str]`: A list of segmentation targets.
```

--------------------------------

### CheXpert Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

CheXpert Dataset, a large chest radiograph dataset with uncertainty labels and expert comparison.

```APIDOC
## CheXpert Dataset

### Description
CheXpert Dataset. A small validation set is provided with the data as well, but is so tiny, it is not included here.

### Citation
CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. Jeremy Irvin *, Pranav Rajpurkar *, Michael Ko, Yifan Yu, Silviana Ciurea-Ilcus, Chris Chute, Henrik Marklund, Behzad Haghgoo, Robyn Ball, Katie Shpanskaya, Jayne Seekins, David A. Mong, Safwan S. Halabi, Jesse K. Sandberg, Ricky Jones, David B. Larson, Curtis P. Langlotz, Bhavik N. Patel, Matthew P. Lungren, Andrew Y. Ng. https://arxiv.org/abs/1901.07031
Dataset website here: https://stanfordmlgroup.github.io/competitions/chexpert/

### Class
`_torchxrayvision.datasets.CheX_Dataset`

### Parameters
- **_imgpath_** (str) - Path to the CheXpert dataset images.
- **_csvpath_** (str, optional) - Path to the CSV file. Defaults to 'USE_INCLUDED_FILE'.
- **_views_** (list, optional) - List of views to include. Defaults to ['PA'].
- **_transform_** (callable, optional) - Optional transform to be applied on a sample.
- **_data_aug_** (callable, optional) - Optional data augmentation to be applied on a sample.
- **_flat_dir_** (bool, optional) - Whether the image directory is flat. Defaults to True.
- **_seed_** (int, optional) - Random seed for reproducibility. Defaults to 0.
- **_unique_patients_** (bool, optional) - Whether to ensure unique patients. Defaults to True.
```

--------------------------------

### Mira Sex Model

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the Mira Sex model.

```APIDOC
## Mira Sex Model

### Class
`torchxrayvision.models.SexModel`

### Description
A model trained on Mira data for predicting the sex of the patient.

**Attributes:**
- `targets: List[str]`: A list of sex predictions.
```

--------------------------------

### Model Interface

Source: https://mlmed.org/torchxrayvision/models.html

Details the core Model interface, including forward pass, targets, and feature extraction.

```APIDOC
## Model Interface

### Description
The `Model` class is the base interface for all classifiers in the library. It handles input image resizing and provides methods for prediction and feature extraction.

### Class
`torchxrayvision.models.Model`

### Methods
#### `forward(_x: Tensor) -> Tensor`
Performs a forward pass through the model. Outputs a tensor with shape `[batch, pathologies]` aligned with `model.pathologies`.

**Example:**
```python
preds = model(img)
print(dict(zip(model.targets, preds.tolist()[0])))
```

#### `targets: List[str]`
Attribute that provides a list of pathology names corresponding to the model's predictions. This list can change depending on the loaded weights.

#### `features(_x: Tensor) -> Tensor`
Extracts feature vectors from the input image. Useful for semi-supervised training or transfer learning.

**Example:**
```python
feats = model.features(img)
```
```

--------------------------------

### Predicting Patient Sex with MIRA Model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the model and performs inference on a loaded chest X-ray image.

```python
model = xrv.baseline_models.mira.SexModel()

image = xrv.utils.load_image('00027426_000.png')
image = torch.from_numpy(image)[None,...]

pred = model(image)

model.targets[torch.argmax(pred)]
# 'Male' or 'Female'  
```

--------------------------------

### MIRA Publication BibTeX

Source: https://mlmed.org/torchxrayvision/models.html

Citation information for the MIRA project.

```bibtex
@article{MIRA2023,
    title = {Chexploration: Medical Image Representation and Analysis},
    author = {MIRA Team},
    journal = {biomedia-mira/chexploration},
    url = {https://github.com/biomedia-mira/chexploration},
    year = {2023}
}
```

--------------------------------

### Class: StonyBrookCOVID_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Initializes the Stonybrook Radiographic Assessment of Lung Opacity Score Dataset loader.

```APIDOC
## StonyBrookCOVID_Dataset

### Description
Loads the Stonybrook Radiographic Assessment of Lung Opacity Score Dataset.

### Parameters
- **imgpath** (str) - Required - Path to the image directory.
- **csvpath** (str) - Required - Path to the CSV file.
- **transform** (callable, optional) - Optional - Transformation function.
- **data_aug** (callable, optional) - Optional - Data augmentation function.
- **views** (list) - Optional - List of views to include (default: ['AP']).
- **seed** (int) - Optional - Random seed (default: 0).
```

--------------------------------

### Dataset Interface

Source: https://mlmed.org/torchxrayvision/datasets.html

Standard interface for interacting with medical imaging datasets in TorchXRayVision.

```APIDOC
## Dataset Interface

### Description
Base class for handling medical imaging datasets, providing access to labels, paths, and metadata.

### Attributes
- **pathologies**: List of pathologies included in the dataset.
- **labels**: The ground truth labels for the dataset.
- **csv**: The underlying metadata CSV file.

### Methods
- **totals()**: Returns the count of samples per pathology.
- **__repr__()**: Returns a string representation of the dataset.
```

--------------------------------

### Xinario View Model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the ViewModel to classify the view of a chest X-ray.

```APIDOC
## ViewModel

### Description
Classifies the view of a chest X-ray. Images are automatically scaled to 320x320.

### Response
- **targets** (list) - ['Frontal', 'Lateral']
```

--------------------------------

### Emory HITI Race Model

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the Emory HITI Race model.

```APIDOC
## Emory HITI Race Model

### Class
`torchxrayvision.models.RaceModel`

### Description
This model is designed for race prediction based on HITI data from Emory.

**Attributes:**
- `targets: List[str]`: A list of race-related predictions.
```

--------------------------------

### XRV Pathology Classifiers

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for specific XRV pathology classifier models like DenseNet and ResNet.

```APIDOC
## XRV Pathology Classifiers

### DenseNet

**Description:**
Based on "Densely Connected Convolutional Networks". This classifier is trained to predict various pathologies.

**Class:**
`torchxrayvision.models.DenseNet`

**Parameters:**
- `_weights` (SPECIFY): Specifies the pre-trained weights to load. (Required)
- `_op_threshs` (Optional): Thresholds for operation.
- `_apply_sigmoid` (Optional): Whether to apply sigmoid activation to the output.

**Attributes:**
- `targets: List[str]`: A list of pathology names that this DenseNet model predicts.
```

```APIDOC
### ResNet

**Description:**
Implementation of the ResNet architecture for pathology classification.

**Class:**
`torchxrayvision.models.ResNet`

**Attributes:**
- `targets: List[str]`: A list of pathology names that this ResNet model predicts.
```

--------------------------------

### CheXpert Pathology Classifier

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the CheXpert pathology classifier.

```APIDOC
## CheXpert Pathology Classifier

### Class
`torchxrayvision.models.DenseNet`

### Description
A DenseNet model specifically trained for the CheXpert dataset to classify pathologies.

**Attributes:**
- `targets: List[str]`: A list of pathology names predicted by the CheXpert classifier.
```

--------------------------------

### Create Dataset Subset by Index

Source: https://mlmed.org/torchxrayvision/dataset_helpers.html

Use SubsetDataset to create a new dataset containing only specific data points identified by their indices. This is useful for creating custom views or samples of your data.

```python
dsubset = xrv.datasets.SubsetDataset(dataset, [0, 5, 60])
# Output:
SubsetDataset num_samples=3
of PC_Dataset num_samples=94825 views=['PA', 'AP']
```

--------------------------------

### Model Interface

Source: https://mlmed.org/torchxrayvision/index.html

The base Model interface for all TorchXRayVision models, providing methods for forward passes and feature extraction.

```APIDOC
## Model Interface

### Description
The base class for all models in the library, defining the standard interface for inference and feature extraction.

### Methods
- **forward()**: Performs the forward pass of the model.
- **features()**: Extracts feature representations from the model.

### Attributes
- **targets**: A list of target labels or pathologies the model is trained to predict.
```

--------------------------------

### NIH_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Interface for the NIH ChestX-ray14 dataset.

```APIDOC
## Class: NIH_Dataset

### Description
Interface for the NIH ChestX-ray14 dataset containing 112,120 frontal-view X-ray images.

### Parameters
- **imgpath** (str) - Required - Path to the images.
- **csvpath** (str) - Optional - Path to the metadata CSV file (default: 'USE_INCLUDED_FILE').
- **bbox_list_path** (str) - Optional - Path to bounding box list (default: 'USE_INCLUDED_FILE').
- **views** (List[str]) - Optional - List of views to include (default: ['PA']).
- **transform** (callable) - Optional - Transform to apply to images.
- **data_aug** (callable) - Optional - Data augmentation to apply.
- **seed** (int) - Optional - Random seed (default: 0).
- **unique_patients** (bool) - Optional - Whether to filter for unique patients (default: True).
- **pathology_masks** (bool) - Optional - Whether to include pathology masks (default: False).
```

--------------------------------

### PadChest Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

PadChest dataset from the Hospital San Juan de Alicante - University of Alicante. Images with null labels or that cannot be loaded are excluded.

```APIDOC
## PadChest Dataset

### Description
PadChest dataset from the Hospital San Juan de Alicante - University of Alicante. Note that images with null labels (as opposed to normal), and images that cannot be properly loaded (listed as ‘missing’ in the code) are excluded, which makes the total number of available images slightly less than the total number of image files.

### Citation
PadChest: A large chest x-ray image dataset with multi-label annotated reports. Aurelia Bustos, Antonio Pertusa, Jose-Maria Salinas, and Maria de la Iglesia-Vayá. arXiv preprint, 2019. https://arxiv.org/abs/1901.07042
Dataset website: http://bimcv.cipf.es/bimcv-projects/padchest/
Download full size images here: https://academictorrents.com/details/dec12db21d57e158f78621f06dcbe78248d14850
Download resized (224x224) images here (recropped): https://academictorrents.com/details/96ebb4f92b85929eadfb16761f310a6d04105797

### Class
`_torchxrayvision.datasets.PC_Dataset`

### Parameters
- **_imgpath_** (str) - Path to the PadChest dataset images.
- **_csvpath_** (str, optional) - Path to the CSV file. Defaults to 'USE_INCLUDED_FILE'.
- **_views_** (list, optional) - List of views to include. Defaults to ['PA'].
- **_transform_** (callable, optional) - Optional transform to be applied on a sample.
- **_data_aug_** (callable, optional) - Optional data augmentation to be applied on a sample.
- **_flat_dir_** (bool, optional) - Whether the image directory is flat. Defaults to True.
- **_seed_** (int, optional) - Random seed for reproducibility. Defaults to 0.
- **_unique_patients_** (bool, optional) - Whether to ensure unique patients. Defaults to True.
```

--------------------------------

### Riken age model citation

Source: https://mlmed.org/torchxrayvision/models.html

BibTeX citation for the Riken age estimation model.

```bibtex
@article{Ieki2022,
    title = {{Deep learning-based age estimation from chest X-rays indicates cardiovascular prognosis}},
    author = {Ieki, Hirotaka et al.},
    doi = {10.1038/s43856-022-00220-6},
    journal = {Communications Medicine},
    publisher = {Nature Publishing Group},
    url = {https://www.nature.com/articles/s43856-022-00220-6},
    year = {2022}
}
```

--------------------------------

### RSNA_Pneumonia_Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Interface for the RSNA Pneumonia Detection Challenge dataset.

```APIDOC
## Class: RSNA_Pneumonia_Dataset

### Description
Interface for the RSNA Pneumonia Detection Challenge dataset.

### Parameters
- **imgpath** (str) - Required - Path to the images.
- **csvpath** (str) - Optional - Path to the metadata CSV file (default: 'USE_INCLUDED_FILE').
- **dicomcsvpath** (str) - Optional - Path to the DICOM metadata CSV (default: 'USE_INCLUDED_FILE').
- **views** (List[str]) - Optional - List of views to include (default: ['PA']).
- **transform** (callable) - Optional - Transform to apply to images.
- **data_aug** (callable) - Optional - Data augmentation to apply.
- **nrows** (int) - Optional - Number of rows to read from CSV.
- **seed** (int) - Optional - Random seed (default: 0).
- **pathology_masks** (bool) - Optional - Whether to include pathology masks (default: False).
- **extension** (str) - Optional - File extension for images (default: '.jpg').
```

--------------------------------

### Dataset Helpers

Source: https://mlmed.org/torchxrayvision/index.html

Utility functions and classes for processing, filtering, and merging medical imaging datasets.

```APIDOC
## Dataset Helpers

### Description
Tools for manipulating datasets, including cropping, resizing, and filtering.

### Classes/Functions
- **MergeDataset**: Combines multiple datasets into one.
- **FilterDataset**: Filters samples based on specific criteria.
- **SubsetDataset**: Creates a subset of an existing dataset.
- **XRayCenterCrop**: Applies center cropping to X-ray images.
- **XRayResizer**: Resizes X-ray images to a target dimension.
- **relabel_dataset()**: Utility function to modify dataset labels.
```

--------------------------------

### Model Prediction and Target Mapping

Source: https://mlmed.org/torchxrayvision/models.html

Use the model to predict pathologies on an image and map the predictions to pathology names using the model's targets.

```python
preds = model(img)
print(dict(zip(model.targets, preds.tolist()[0])))
# {'Atelectasis': 0.5583771,
#  'Consolidation': 0.5279943,
#  'Infiltration': 0.60061914,
#  ...
```

--------------------------------

### Use ResNet Autoencoder

Source: https://mlmed.org/torchxrayvision/models.html

Perform encoding and decoding operations using the ResNetAE autoencoder.

```python
ae = xrv.autoencoders.ResNetAE(weights="101-elastic") # trained on PadChest, NIH, CheXpert, and MIMIC
z = ae.encode(image)
image2 = ae.decode(z)
```

--------------------------------

### ChestX-Det citation

Source: https://mlmed.org/torchxrayvision/models.html

BibTeX citation for the ChestX-Det model.

```bibtex
@article{Lian2021,
    title = {{A Structure-Aware Relation Network for Thoracic Diseases Detection and Segmentation}},
    author = {Lian, Jie and Liu, Jingyu and Zhang, Shu and Gao, Kai and Liu, Xiaoqing and Zhang, Dingwen and Yu, Yizhou},
    doi = {10.48550/arxiv.2104.10326},
    journal = {IEEE Transactions on Medical Imaging},
    url = {https://arxiv.org/abs/2104.10326},
    year = {2021}
}
```

--------------------------------

### MIMIC-CXR Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

MIMIC-CXR Dataset, a large publicly available database of labeled chest radiographs.

```APIDOC
## MIMIC-CXR Dataset

### Description
MIMIC-CXR Dataset.

### Citation
Johnson AE, Pollard TJ, Berkowitz S, Greenbaum NR, Lungren MP, Deng CY, Mark RG, Horng S. MIMIC-CXR: A large publicly available database of labeled chest radiographs. arXiv preprint arXiv:1901.07042. 2019 Jan 21.
https://arxiv.org/abs/1901.07042
Dataset website here: https://physionet.org/content/mimic-cxr-jpg/2.0.0/

### Class
`_torchxrayvision.datasets.MIMIC_Dataset`

### Parameters
- **_imgpath_** (str) - Path to the MIMIC-CXR dataset images.
- **_csvpath_** (str) - Path to the CSV file containing image labels.
- **_metacsvpath_** (str) - Path to the CSV file containing metadata.
- **_views_** (list, optional) - List of views to include. Defaults to ['PA'].
- **_transform_** (callable, optional) - Optional transform to be applied on a sample.
- **_data_aug_** (callable, optional) - Optional data augmentation to be applied on a sample.
- **_seed_** (int, optional) - Random seed for reproducibility. Defaults to 0.
- **_unique_patients_** (bool, optional) - Whether to ensure unique patients. Defaults to True.
```

--------------------------------

### Create Dataset Subset by Condition

Source: https://mlmed.org/torchxrayvision/dataset_helpers.html

Use SubsetDataset with indices derived from conditions (e.g., patient sex) to filter datasets. This allows for creating subsets based on specific data attributes.

```python
idxs = np.where(dataset.csv.PatientSex_DICOM=="F")[0]
dsubset = xrv.datasets.SubsetDataset(dataset, idxs)
```

--------------------------------

### Load and run Emory HITI race prediction model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the RaceModel and predicts patient race from a chest X-ray image. Images are automatically scaled to the model's native 320x320 resolution.

```python
model = xrv.baseline_models.emory_hiti.RaceModel()

image = xrv.utils.load_image('00027426_000.png')
image = torch.from_numpy(image)[None,...]

pred = model(image)

model.targets[torch.argmax(pred)]
# 'White'
```

--------------------------------

### OpenI Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

OpenI Dataset, a collection of radiology examinations prepared for distribution and retrieval.

```APIDOC
## OpenI Dataset

### Description
OpenI Dataset. Views have been determined by projection using T-SNE. To use the T-SNE view rather than the view defined by the record, set use_tsne_derived_view to true.

### Citation
Dina Demner-Fushman, Marc D. Kohli, Marc B. Rosenman, Sonya E. Shooshan, Laritza Rodriguez, Sameer Antani, George R. Thoma, and Clement J. McDonald. Preparing a collection of radiology examinations for distribution and retrieval. Journal of the American Medical Informatics Association, 2016. doi: 10.1093/jamia/ocv080.
Dataset website: https://openi.nlm.nih.gov/faq
Download images: https://academictorrents.com/details/5a3a439df24931f410fac269b87b050203d9467d

### Class
`_torchxrayvision.datasets.Openi_Dataset`

### Parameters
- **_imgpath_** (str) - Path to the OpenI dataset images.
- **_xmlpath_** (str, optional) - Path to the XML file. Defaults to 'USE_INCLUDED_FILE'.
- **_dicomcsv_path_** (str, optional) - Path to the DICOM CSV file. Defaults to 'USE_INCLUDED_FILE'.
- **_tsnepacsv_path_** (str, optional) - Path to the t-SNE derived CSV file. Defaults to 'USE_INCLUDED_FILE'.
- **_use_tsne_derived_view_** (bool, optional) - Whether to use t-SNE derived views. Defaults to False.
- **_views_** (list, optional) - List of views to include. Defaults to ['PA'].
- **_transform_** (callable, optional) - Optional transform to be applied on a sample.
- **_data_aug_** (callable, optional) - Optional data augmentation to be applied on a sample.
- **_nrows_** (int, optional) - Number of rows to load from the CSV files.
- **_seed_** (int, optional) - Random seed for reproducibility. Defaults to 0.
- **_unique_patients_** (bool, optional) - Whether to ensure unique patients. Defaults to True.
```

--------------------------------

### CheXpert Pathology Classifier (DenseNet)

Source: https://mlmed.org/torchxrayvision/models.html

A DenseNet model specifically trained for the CheXpert dataset, designed for pathology classification.

```APIDOC
## CheXpert Pathology Classifier

### Description
A DenseNet model specifically trained for the CheXpert dataset, designed for pathology classification. It can load a subset of the ensemble by setting `num_models` less than 30.

### Method
`xrv.baseline_models.chexpert.DenseNet(weights_zip, num_models)`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
# This is a conceptual example as the weights need to be downloaded separately.
# import torchxrayvision as xrv
# model = xrv.baseline_models.chexpert.DenseNet(weights_zip='path/to/chexpert_weights.zip', num_models=30)
```

### Parameters
* **weights_zip** (str) - Path to the zip file containing the weights.
* **num_models** (int) - Number of models to load from the ensemble (default is 30).

### Targets
['Atelectasis', 'Cardiomegaly', 'Consolidation', 'Edema', 'Effusion']
```

--------------------------------

### Simulate Covariate Shift with CovariateDataset

Source: https://mlmed.org/torchxrayvision/dataset_helpers.html

Use CovariateDataset to combine two datasets with a specified ratio to simulate covariate shifts between training and test distributions.

```python
d = xrv.datasets.CovariateDataset(
    d1 = # dataset1 with a specific condition.
    d1_target = # target label to predict.
    d2 = # dataset2 with a specific condition.
    d2_target = #target label to predict.
    mode="train", # train, valid, or test.
    ratio=0.75
)
```

--------------------------------

### Emory HITI Race Model

Source: https://mlmed.org/torchxrayvision/models.html

Initializes the RaceModel to predict patient race from chest X-rays.

```APIDOC
## RaceModel

### Description
Predicts patient race from a chest X-ray using models trained on the MIMIC dataset. Images are automatically scaled to 320x320.

### Response
- **targets** (list) - ['Asian', 'Black', 'White']
```

--------------------------------

### Feature Extraction with Model

Source: https://mlmed.org/torchxrayvision/models.html

Extract feature vectors from images using the model's features method for semi-supervised or transfer learning tasks.

```python
feats = model.features(img)
```

--------------------------------

### XRV ResNet Autoencoder

Source: https://mlmed.org/torchxrayvision/models.html

Utilize a ResNet-based autoencoder for image encoding and decoding. This model is trained on a combination of datasets.

```APIDOC
## XRV ResNet Autoencoder

### Description
A ResNet based autoencoder. This model can be used for image encoding and decoding tasks.

### Method
`xrv.autoencoders.ResNetAE(weights, cache_dir)`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
import torchxrayvision as xrv

# Instantiate the autoencoder with specific weights
ae = xrv.autoencoders.ResNetAE(weights="101-elastic") # trained on PadChest, NIH, CheXpert, and MIMIC

# Encode an image
z = ae.encode(image)

# Decode the encoded representation
image2 = ae.decode(z)
```

### Parameters
* **weights** (str) - Weights to use. See above for options.
* **cache_dir** (str) - Override directory used to store cached weights (default: ~/.torchxrayvision/)
```

--------------------------------

### NIH Dataset

Source: https://mlmed.org/torchxrayvision/datasets.html

Relabelling of a subset of images from the NIH dataset. Data tables should be applied against an NIH download. A test and validation split are provided in the original. They are combined here, but one or the other can be used by providing the original csv to the csvpath argument.

```APIDOC
## NIH Dataset

### Description
A relabelling of a subset of images from the NIH dataset. Data tables should be applied against an NIH download. A test and validation split are provided in the original. They are combined here, but one or the other can be used by providing the original csv to the csvpath argument.

### Citation
Chest Radiograph Interpretation with Deep Learning Models: Assessment with Radiologist-adjudicated Reference Standards and Population-adjusted Evaluation Anna Majkowska, Sid Mittal, David F. Steiner, Joshua J. Reicher, Scott Mayer McKinney, Gavin E. Duggan, Krish Eswaran, Po-Hsuan Cameron Chen, Yun Liu, Sreenivasa Raju Kalidindi, Alexander Ding, Greg S. Corrado, Daniel Tse, and Shravya Shetty. Radiology 2020
https://pubs.rsna.org/doi/10.1148/radiol.2019191293
NIH data can be downloaded here: https://academictorrents.com/details/e615d3aebce373f1dc8bd9d11064da55bdadede0

### Class
`_torchxrayvision.datasets.NIH_Dataset`

### Parameters
- **_imgpath_** (str) - Path to the NIH dataset images.
- **_csvpath_** (str, optional) - Path to the CSV file. Defaults to 'USE_INCLUDED_FILE'.
- **_views_** (list, optional) - List of views to include. Defaults to ['PA'].
- **_transform_** (callable, optional) - Optional transform to be applied on a sample.
- **_data_aug_** (callable, optional) - Optional data augmentation to be applied on a sample.
- **_flat_dir_** (bool, optional) - Whether the image directory is flat. Defaults to True.
- **_seed_** (int, optional) - Random seed for reproducibility. Defaults to 0.
- **_unique_patients_** (bool, optional) - Whether to ensure unique patients. Defaults to True.
```

--------------------------------

### Emory HITI citation

Source: https://mlmed.org/torchxrayvision/models.html

BibTeX citation for the Emory HITI race model.

```bibtex
@article{Gichoya2022,
    title = {AI recognition of patient race in medical imaging: a modelling study},
    author = {Gichoya, Judy Wawira and Banerjee, Imon and Bhimireddy, Ananth Reddy and Burns, John L and Celi, Leo Anthony and Chen, Li-Ching and Correa, Ramon and Dullerud, Natalie and Ghassemi, Marzyeh and Huang, Shih-Cheng and Kuo, Po-Chih and Lungren, Matthew P and Palmer, Lyle J and Price, Brandon J and Purkayastha, Saptarshi and Pyrros, Ayis T and Oakden-Rayner, Lauren and Okechukwu, Chima and Seyyed-Kalantari, Laleh and Trivedi, Hari and Wang, Ryan and Zaiman, Zachary and Zhang, Haoran},
    doi = {10.1016/S2589-7500(22)00063-2},
    journal = {The Lancet Digital Health},
    pmid = {35568690},
    url = {https://www.thelancet.com/journals/landig/article/PIIS2589-7500(22)00063-2/fulltext},
    year = {2022}
}
```

--------------------------------

### JF Healthcare Pathology Classifier

Source: https://mlmed.org/torchxrayvision/models.html

Documentation for the JF Healthcare pathology classifier.

```APIDOC
## JF Healthcare Pathology Classifier

### Class
`torchxrayvision.models.DenseNet`

### Description
A DenseNet model trained on JF Healthcare data for pathology classification.

**Attributes:**
- `targets: List[str]`: A list of pathology names predicted by the JF Healthcare classifier.
```
Tokens:
Raw