rad-dino-maira-2

Model card for RAD-DINO-MAIRA-2

<!-- Provide a quick summary of what the model is/does. -->

Model description

<!-- Provide a longer summary of what this model is. -->

RAD-DINO-MAIRA-2 is a vision transformer model trained to encode chest X-rays using the self-supervised learning method DINOv2. RAD-DINO-MAIRA-2 is a variant of RAD-DINO, which is described in detail in RAD-DINO: Exploring Scalable Medical Image Encoders Beyond Text Supervision (F. Pérez-García, H. Sharma, S. Bond-Taylor, et al., 2024).

RAD-DINO-MAIRA-2 is the version of RAD-DINO used in MAIRA-2: Grounded Radiology Report Generation (S. Bannur, K. Bouzid, et al., 2024). Relative to RAD-DINO, it was trained on more data.

• Developed by: Microsoft Health Futures
• Model type: Vision transformer
• License: MSRLA
• Finetuned from model: dinov2-base

Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

RAD-DINO-MAIRA-2 is shared for research purposes only. It is not meant to be used for clinical practice.

<!-- ### Downstream use --> <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->

The model is a vision backbone that can be plugged to other models for downstream tasks. Some potential uses are:

• Image classification, with a classifier trained on top of the CLS token
• Image segmentation, with a decoder trained using the patch tokens
• Clustering, using the image embeddings directly
• Image retrieval, using nearest neighbors of the CLS token
• Report generation, with a language model to decode text

Fine-tuning RAD-DINO-MAIRA-2 is typically not necessary to obtain good performance in downstream tasks.

<!-- ### Out-of-scope use --> <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

Biases, risks, and limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

RAD-DINO-MAIRA-2 was trained with data from three countries, therefore it might be biased towards population in the training data. Underlying biases of the training datasets may not be well characterized.

Getting started

from transformers import pipeline
pipe = pipeline(task="image-feature-extraction", model="microsoft/rad-dino-maira-2", pool=False)
patch_features = pipe("https://www.bhf.org.uk/-/media/images/information-support/tests/chest-x-ray/normal-chest-x-ray-620x400.jpg")

Refer to RAD-DINO for a more detailed example.

Training details

Training data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

We used images from five public and one private deidentified chest X-ray datasets to train RAD-DINO-MAIRA-2.

Images in the validation and test sets used to train MAIRA-2 were excluded from the training set of RAD-DINO-MAIRA-2.

We used 8 nodes with 4 A100 GPUs each, and a batch size of 40 images per GPU. We share the last checkpoint, trained for 105 000 steps.

Training procedure

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

We refer to the manuscript for a detailed description of the training procedure.

Preprocessing

All DICOM files were resized using B-spline interpolation so that their shorter size was 518, min-max scaled to [0, 255], and stored as PNG files.

Training hyperparameters

• Training regime: fp16 using PyTorch-FSDP mixed-precision.

<!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

Our evaluation is best described in the manuscript.

<!-- ### Testing data, factors & metrics #### Testing Data [More Information Needed] #### Factors [More Information Needed] #### Metrics [More Information Needed] ### Results [More Information Needed] #### Summary -->

Environmental impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> <!-- Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). --> <!-- Hardware type: A100 PCIe --> <!-- Hours: 1d 17h = 41h --> <!-- Cloud provider: Azure --> <!-- Region: West US 2 -->

• Hardware type: NVIDIA A100 GPUs
• Hours used: 41 hours/GPU × 8 nodes × 4 GPUs/node = 1312 GPU-hours
• Cloud provider: Azure
• Compute region: West US 2
• Carbon emitted: 98.4 kg CO₂ eq.

Compute infrastructure

RAD-DINO-MAIRA-2 was trained on Azure Machine Learning.

Hardware

We used 8 Standard_NC96ads_A100_v4 nodes with four NVIDIA A100 (80 GB) GPUs each.

Software

We leveraged the code in DINOv2 for training. We used SimpleITK and Pydicom for processing of DICOM files.

Citation

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

BibTeX:

@misc{perezgarcia2024raddino,
      title={{RAD-DINO}: Exploring Scalable Medical Image Encoders Beyond Text Supervision},
      author={Fernando Pérez-García and Harshita Sharma and Sam Bond-Taylor and Kenza Bouzid and Valentina Salvatelli and Maximilian Ilse and Shruthi Bannur and Daniel C. Castro and Anton Schwaighofer and Matthew P. Lungren and Maria Wetscherek and Noel Codella and Stephanie L. Hyland and Javier Alvarez-Valle and Ozan Oktay},
      year={2024},
      eprint={2401.10815},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

APA:

Pérez-García, F., Sharma, H., Bond-Taylor, S., Bouzid, K., Salvatelli, V., Ilse, M., Bannur, S., Castro, D.C., Schwaighofer, A., Lungren, M.P., Wetscherek, M.T., Codella, N., Hyland, S.L., Alvarez-Valle, J., & Oktay, O. (2024). RAD-DINO: Exploring Scalable Medical Image Encoders Beyond Text Supervision. ArXiv, abs/2401.10815.

Model card contact

Fernando Pérez-García ([email protected]).