SIIM-ACR-Pneumothorax

This dataset features 2,669 high-resolution chest X-ray images (1024×1024) with corresponding lung segmentation masks, specifically curated for pneumothorax detection and medical image segmentation. Modified from the original dataset, it ensures only valid image-mask pairs, making it ideal for deep learning applications in medical AI, radiology, and healthcare research.

Dataset Characteristics

• Total Images (Chest X-rays): 2,669
• Total Masks (Lung Segmentation): 2,669
• Training Split: 2,379 images & masks
• Testing Split: 290 images & masks
• Dataset Size: 1.38 GB

Image & Mask Properties

• X-ray Images:
  • Size: 1024×1024
  • Mode: RGB (Converted from grayscale)
  • Channels: 3
• Segmentation Masks:
  • Size: 1024×1024
  • Mode: Grayscale
  • Channels: 1

Modifications & Enhancements

✅ Filtered Valid Image-Mask Pairs – Removed empty or irrelevant masks.
✅ Grayscale to RGB Conversion – Enhanced image compatibility for deep learning models.

Applications

✅ Pneumothorax Detection & Diagnosis 🩺
✅ Medical Image Segmentation & AI Training 🤖
✅ Deep Learning & Radiology Research 🏥

This dataset is sourced from Kaggle.