Open AccessClassifying HCM Subtypes via Left Ventricular Two-Dimensional Curvature and Dynamic Time Wrapping
Author(s) -
Min Wan,
Yuli Yang,
Xiaodan Zhao,
Shuang Leng,
Jun-Mei Zhang,
Ru San Tan,
Liang Zhong
Publication year - 2020
Publication title -
2019 computing in cardiology (cinc)
Language(s) - English
Resource type - Conference proceedings
SCImago Journal Rank - 0.257
H-Index - 55
ISSN - 2325-887X
ISBN - 978-1-7281-6936-1
DOI - 10.22489/cinc.2019.188
Subject(s) - bioengineering , computing and processing , signal processing and analysis
Identification of Hypertrophic cardiomyopathy (HCM) morphological subtypes has the same significance as diagnosis HCM due to the close relation of subtypes and genetic mutations. This study aims to automatically classify HCM cases into subtypes via two-dimensional endocardial curvature and dynamic time wrapping (DTW). A total of 19 HCM patients underwent cardiac magnetic resonance (CMR) scans. Endocardial boundaries were delineated on long axis images. Endocardial curvatures were computed. Dynamic time wrapping and its variant soft-DTW were used to quantify the dissimilarity between a pair of endocardial curvature series. The dissimilarity was utlized in classifiers. The classifier using DTW has a 0.84 in-sample accuracy and 0.58 out-of-sample accuracy. Soft-DTW has a 0.79 in-sample accuracy and 0.68 out-of-sample accuracy. We envision the use of this method in current clinical practice to improve the subtype identification efficiency and effectiveness.