Discrimination models with radiomics features derived from cardiovascular magnetic resonance images for...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Differentiating hypertrophic cardiomyopathy (HCM) and hypertensive heart disease (HHD) is clinically challenging due to shared feature of left ventricular hypertrophy (LVH).
• Accurate discrimination is crucial for appropriate patient management and treatment strategies.

Purpose of the Study:

• To investigate radiomics features from cardiac magnetic resonance (CMR) imaging for differentiating HHD from HCM.
• To develop and validate discrimination models incorporating radiomics scores (radscore) and multiparametric CMR findings.

Main Methods:

• A retrospective study included 421 HCM and 200 HHD patients, randomly stratified into training (60%) and validation (40%) groups.
• Multiparametric CMR findings and radiomics features were extracted from late gadolinium enhancement (LGE) and cine images.
• Discrimination models were built using logistic regression, combining CMR findings and radscores.

Main Results:

• Maximal left ventricular end-diastolic wall thickness, left ventricular ejection fraction, LGE presence, cine radscore, and LGE radscore were significant discriminators.
• The combined model achieved high diagnostic performance with an area under the receiver operator characteristic curve (AUC) of 0.979 in the training group and 0.981 in the validation group.
• The combined model significantly outperformed models using only multiparametric CMR findings (P<0.001).

Conclusions:

• Radiomics features derived from cardiac cine and LGE images effectively discriminate between HHD and HCM.
• The developed radiomics-based model offers a promising tool for improving the diagnostic accuracy in differentiating these two conditions.