M-FLAG Checkpoint Model Card

A retrained M-FLAG model for benchmarking medical vision-language pre-training methods within the BenchX framework.

Model Details

• Model Type: M-FLAG
• Architecture: ResNet-50 image encoder and CXR-BERT text encoder
• Original Papers: M-FLAG: Medical Vision-Language Pre-training with Frozen Language Models and Latent Space Geometry Optimization
• Benchmark Paper: BenchX: A Unified Benchmark Framework for Medical Vision-Language Pretraining on Chest X-Rays
• Benchmark Framework: https://github.com/yangzhou12/BenchX

Intended Use

• Primary Use Cases:
  • Benchmarking performance for Medical Image Classification
  • Benchmarking performance for Medical Image Segmentation
  • Benchmarking performance for Medical Report Generation

Pre-Training Data

• Dataset:
  • Data source(s): MIMIC-CXR
  • Types of medical images: Frontal chest X-rays
  • Text data type: Associated radiology reports

Prerequisites

Please follow the instruction to install BenchX.

Training & Evaluation

1. Classification

To fine-tune M-FLAG for classification, run this command:

python bin/train.py config/classification/<dataset_name>/mflag.yml

2. Segmentation

To fine-tune M-FLAG for segmentation, run this command:

python mmsegmentation/tools/train.py config/benchmark/<dataset_name>/mflag.yml

3. Report Generation

To fine-tune M-FLAG for report generation, run this command:

python bin/train.py config/report_generation/<dataset_name>/mflag.yml

4. Evaluation

To evaluate fine-tuned M-FLAG models, run:

# For classification and report generation
python bin/test.py config/<task_name>/<dataset_name>/mflag.yml validator.splits=[test] ckpt_dir=<path_to_checkpoint>

# For segmentation
python mmsegmentation/tools/my_test.py mmsegmentation/config/<dataset_name>/mflag.yml <path_to_checkpoint>

Citations

@inproceedings{huang2021M-FLAG,
  title={M-FLAG: Medical Vision-Language Pre-training with Frozen Language Models and Latent Space Geometry Optimization},
  author={Liu, Che and Cheng, Sibo and Chen, Chen and Qiao, Mengyun and Zhang, Weitong and Shah, Anand and Bai, Wenjia and Arcucci, Rossella},
  booktitle={Proceedings of MICCAI},
  pages={637--647},
  year={2023},
}

@inproceedings{zhou2024benchx,
  title={BenchX: A Unified Benchmark Framework for Medical Vision-Language Pretraining on Chest X-Rays},
  author={Yang Zhou, Tan Li Hui Faith, Yanyu Xu, Sicong Leng, Xinxing Xu, Yong Liu, Rick Siow Mong Goh},
  booktitle={Proceedings of NeurIPS},
  year={2024}
}