FG-CXR: Radiologist-Aligned Gaze Data for Chest X-Ray Report Generation

A multimodal dataset linking chest X-ray images with radiologist gaze patterns and report ground truth. The data includes fixation lists, static gaze heatmaps, segmentation masks, and a full JSON file with report annotations. It was created by phamtrongthang and last updated on Hugging Face in November 2025.

Use Cases

• Training AI models for chest X-ray report generation based on image and report ground truth.
• Studying radiologist visual attention patterns based on eye-tracking fixation lists and heatmaps.
• Developing interpretable AI systems that align with clinical reasoning based on gaze-aligned annotations.
• Benchmarking segmentation models for chest X-ray anatomy based on provided segmentation masks.

Strengths

• Includes multiple aligned data modalities: images, gaze data, segmentation masks, and textual reports.
• Specifically designed for enhancing interpretability in medical AI, as stated in the description.
• Contains ground truth radiology reports for supervised learning tasks.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset scale are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Likely collected from radiologist studies, as it includes eye-tracking data aligned with medical images.

Freshness

Last updated 2025-11-20 18:52:46; freshness should be verified.