Machine learning-based interpretation of non-contrast feature tracking strain analysis and T1/T2 mapping for assessing myocardial viability

• Published in: Scientific reports Vol. 15; no. 1; pp. 753 - 18
• Main Authors: GhaffariJolfayi, Amir, Salmanipour, Alireza, Heshmat-Ghahdarijani, Kiyan, MozafaryBazargany, MohammadHossein, Azimi, Amir, Pirouzi, Pirouz, Mohammadzadeh, Ali
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 04.01.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 692/4019/2776; 692/4019/592; Aged; Algorithms; Cardiovascular diseases; Cardiovascular magnetic resonance; Contrast Media; Coronary artery; Coronary vessels; Data acquisition; Feature tracking strain analysis; Female; Gadolinium; Heart diseases; Humanities and Social Sciences; Humans; Image Interpretation, Computer-Assisted - methods; Ischemia; Learning algorithms; Machine Learning; Magnetic Resonance Imaging - methods; Male; Mapping; Middle Aged; multidisciplinary; Myocardial infarction; Myocardial Infarction - diagnostic imaging; Myocardial viability; Myocardium - pathology; Neural networks; Population studies; Regression analysis; Renal function; Retrospective Studies; Science; Science (multidisciplinary); T1/T2 mapping; Viability; T1/T2 mapping; Myocardial viability; Cardiovascular magnetic resonance; Feature tracking strain analysis
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-85029-0

Assessing myocardial viability is crucial for managing ischemic heart disease. While late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) is the gold standard for viability evaluation, it has limitations, including contraindications in patients with renal dysfunction and lengthy scan times. This study investigates the potential of non-contrast CMR techniques—feature tracking strain analysis and T1/T2 mapping—combined with machine learning (ML) models, as an alternative to LGE-CMR for myocardial viability assessment. A retrospective analysis was conducted on 79 patients with myocardial infarction (MI) 2–4 weeks post-event. Patients with prior ischemia or poor imaging quality were excluded to ensure robust data acquisition. Various ML algorithms were applied to data from LGE-CMR and non-contrast CMR techniques. Random forest (RF) demonstrated the highest predictive accuracy, with area under the curve (AUC) values of 0.89, 0.90, and 0.92 for left anterior descending (LAD), right coronary artery (RCA), and left circumflex (LCX) coronary artery territories, respectively. For the LAD territory, RF, k-nearest neighbors (KNN), and logistic regression were the top performers, while RCA showed the best results from RF, neural networks (NN), and KNN. In the LCX territory, RF, NN, and logistic regression were most effective. The integration of T1/T2 mapping and strain analysis significantly enhanced myocardial viability prediction, positioning these non-contrast techniques as promising alternatives to LGE-CMR. ML models, particularly RF, provided superior diagnostic accuracy across coronary territories. Future studies should validate these findings across diverse populations and clinical settings.