cimti

cimti applies a speckle-tracking strain-based artificial neural network to differentiate ischemic from non-ischemic cardiomyopathy using echocardiographic left ventricular strain parameters for etiologic classification.

Key Features:

• Speckle-tracking strain input: Uses echocardiographic left ventricular strain parameters as the primary input for classification.
• Artificial Neural Network (ANN) model: Implements an ANN trained to distinguish ischemic versus non-ischemic cardiomyopathy based on left ventricular strain features.
• Feature integration: Incorporates regional wall motion abnormalities, electrocardiographic (ECG) features, and demographic information alongside strain parameters.
• Comparator models: Logistic regression models were developed alongside ANN models for performance comparison.
• Validation and performance: Evaluated using derivation and independent validation cohorts, with the strain-based ANN achieving an F1 score of 0.82 versus 0.79 for a full-feature ANN and 0.63 for logistic regression.
• Study cohort: Developed from a retrospective dataset of 204 patients with reduced ejection fraction (<50%) who underwent diagnostic angiography.

Scientific Applications:

• Clinical decision support: Provides non-invasive classification of cardiomyopathy etiology (ischemic vs non-ischemic) to inform clinical management decisions.
• Research tool: Serves as a model-based resource for studies investigating heart failure etiology and the predictive value of strain and other cardiac features.

Methodology:

Retrospective collection of echocardiographic strain parameters and clinical data from 204 patients with reduced EF (<50%) who underwent diagnostic angiography; construction of logistic regression and ANN models in a derivation cohort; validation of model performance in an independent validation cohort.