Unveiling The Patterns Of Cardiac Fibrosis Using 3d Radiomics In Heart Failure Patients

Abstract

Background: Myocardial perfusion imaging using 99mTc-MIBI SPECT is valuable for assessing chronic myocardial infarction (MI). However, conventional analysis may overlook heterogeneity within fixed perfusion deficits.

Objective: This study aims to utilize radionics analysis to identify and characterize heterogeneity within fixed perfusion deficits caused by chronic MI.

Methods: A retrospective cohort study was conducted in a single-centre outpatient setting. Twenty-eight patients (9 women, 9 men; mean age 68 ± 11 years) with fixed perfusion abnormalities on resting polar map images and a history of chronic MI were included. Radiomics data were extracted using Mazda software from regions of interest within the homogeneous-uptake myocardium and fixed defects with probable heterogeneous uptake. Random Forest/LogitBoost algorithms in the WEKA artificial intelligence program were employed for analysis, achieving 100% sensitivity, specificity, and area under the ROC curve (AUC) for classifying heterogeneous vs. homogeneous textures. Heterogeneous textures derived from co-occurrence and run-length matrices were identified as significantly more predictive (AUC 100%) than homogeneous textures by binomial logistic regression analysis.

Results: Identified heterogeneous textures (e.g., 135dr_ShrtREmp, 45dgr_ShrtREmp, 45dgr_Fraction, S(2,-2) InvDfMom, Horzl_Fraction) demonstrated superior predictive capability for heterogeneity within fixed perfusion deficits compared to homogeneous textures.

Conclusion: Radiomics analysis enables the prediction and quantification of heterogeneity within permanent perfusion abnormalities. This approach offers potential future applications for myocardial perfusion SPECT in individuals with persistent myocardial infarction, thereby enhancing diagnostic and prognostic capabilities in this population.