Automated identification and quantification of myocardial inflammatory infiltration in digital histological images to diagnose myocarditis

Abstract

This study aims to develop a new computational pathology approach that automates the identification and quantification of myocardial inflammatory infiltration in digital HE-stained images to provide a quantitative histological diagnosis of myocarditis.898 HE-stained whole slide images (WSIs) of myocardium from 154 heart transplant patients diagnosed with myocarditis or dilated cardiomyopathy (DCM) were included in this study. An automated DL-based computational pathology approach was developed to identify nuclei and detect myocardial inflammatory infiltration, enabling the quantification of the lymphocyte nuclear density (LND) on myocardial WSIs. A cutoff value based on the quantification of LND was proposed to determine if the myocardial inflammatory infiltration was present. The performance of our approach was evaluated with a five-fold cross-validation experiment, tested with an internal test set from the myocarditis group, and confirmed by an external test from a double-blind trial group. An LND of 1.02/mm2 could distinguish WSIs with myocarditis from those without. The accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) in the five-fold cross-validation experiment were 0.899 plus or minus 0.035, 0.971 plus or minus 0.017, 0.728 plus or minus 0.073 and 0.849 plus or minus 0.044, respectively. For the internal test set, the accuracy, sensitivity, specificity, and AUC were 0.887, 0.971, 0.737, and 0.854, respectively. The accuracy, sensitivity, specificity, and AUC for the external test set reached 0.853, 0.846, 0.858, and 0.852, respectively. Our new approach provides accurate and reliable quantification of the LND of myocardial WSIs, facilitating automated quantitative diagnosis of myocarditis with HE-stained images.