Related Experiment Video
Updated: Jun 20, 2025
Assessment of Cardiac Function and Myocardial Morphology Using Small Animal Look-locker Inversion Recovery SALLI MRI in Rats
Published on: July 19, 2013
Enabling Reliable Visual Detection of Chronic Myocardial Infarction with Native T1 Cardiac MRI Using Data-Driven
Khalid Youssef1, Xinheng Zhang1, Ghazal Yoosefian1
1From the Krannert Cardiovascular Research Center, Indiana University School of Medicine, IU Health Cardiovascular Institute, 1700 N Capitol Ave, E316, Indianapolis, IN 46202-1228 (K.Y., X.Z., G.Y., S.F.C., K.V., B.S., R.D.); University of California Los Angeles, Los Angeles, Calif (X.Z.); Zhongshan Hospital, Fudan University, Shanghai, China (Y.C.); Cedars-Sinai Medical Center, Los Angeles, Calif (H.J.Y.); Libin Cardiovascular Institute of Alberta, University of Calgary, Alberta, Canada (A.H.); and Northern Ontario School of Medicine University, Sudbury, Canada (A.K.).
A novel data-driven native mapping (DNM) approach using machine learning significantly enhances contrast in cardiac MRI for visualizing chronic myocardial infarction. This method improves infarct detection compared to standard T1 maps.
Area of Science:
- Cardiovascular Imaging
- Machine Learning in Medicine
- Radiology
Background:
- Chronic myocardial infarction assessment relies on cardiac MRI, often using T1 mapping.
- Standard T1 mapping algorithms may not fully optimize contrast between infarcted and remote myocardial tissues.
- Histologic heterogeneity in infarcts presents challenges for accurate imaging.
Purpose of the Study:
- To evaluate a data-driven native mapping (DNM) approach using machine learning to optimize infarct-to-remote myocardial contrast.
- To replace generic fitting algorithms with a pixel-wise, data-driven method for native T1 maps.
- To improve visualization of chronic reperfused infarcts in a canine model.
Main Methods:
- Utilized a canine model of chronic myocardial infarction (n=24).
- Employed unsupervised clustering (self-organizing maps, t-SNE) for native T1-weighted pixel patterns.
- Trained deep neural networks to map native T1-weighted patterns to late gadolinium enhancement (LGE) images for DNM.
Main Results:
- Distinct pixel-intensity patterns were observed between infarcted and remote territories.
- DNM showed significantly higher contrast-to-noise ratio (15.01 ± 2.88) compared to standard native T1 maps (5.64 ± 1.58; P < .001).
- DNM demonstrated stronger correlation with LGE (R² = 0.85) than native T1 maps (R² = 0.71) for infarct identification.
Conclusions:
- Native T1-weighted pixels contain extractable information for enhancing infarct visualization.
- The proposed DNM approach effectively maximizes image contrast between infarct and remote territories.
- DNM offers a promising non-contrast method for improved chronic infarct characterization in cardiac MRI.
More Related Videos
06:29Cardiac Magnetic Resonance for the Evaluation of Suspected Cardiac Thrombus: Conventional and Emerging Techniques
Published on: June 11, 2019
08:13In Vivo Quantitative Assessment of Myocardial Structure, Function, Perfusion and Viability Using Cardiac Micro-computed Tomography
Published on: February 16, 2016
