HMC-QU Dataset

The HMC-QU Dataset is a benchmark resource for echocardiography-based myocardial infarction (MI) detection and left ventricle wall segmentation. Developed collaboratively by Hamad Medical Corporation (HMC), Tampere University, and Qatar University, this dataset is ethically approved and offers a rich collection of high-resolution echocardiographic recordings. It supports researchers and practitioners working on cardiac health diagnostics, machine learning models, and deep learning applications.

Dataset Highlights

• Echocardiography Views:
  • 162 recordings in Apical 4-Chamber (A4C) view.
  • 160 recordings in Apical 2-Chamber (A2C) view.
• Ground-Truth Labels: Includes segment-level annotations for MI and non-MI based on regional wall motion abnormalities.
• Temporal Resolution: Recordings at 25 fps.
• Spatial Resolution: 422×636 to 768×1024 pixels.
• Device Diversity: Data collected using Philips and GE Vivid ultrasound machines.

Key Features

1. Myocardial Infarction Detection

   • The dataset includes over 10,000 echocardiograms, with 800+ cases of acute ST-elevation myocardial infarction (STEMI).
   • MI cases were treated with coronary angioplasty post-diagnosis using electrocardiography (ECG) and cardiac enzyme evidence.
   • Ground-truth labels define MI and non-MI segments based on regional wall motion abnormalities.

2. Left Ventricle Wall Segmentation

   • A subset of 109 A4C recordings includes segmentation masks for the entire left ventricle wall.
   • Ground-truth masks are resized to 224×224 pixels for compatibility with deep learning models.
   • The subset features 72 MI patients and 37 non-MI subjects.