Short-axis cine cardiac magnetic resonance images-derived radiomics for hypertrophic cardiomyopathy and healthy control classification

doi:10.3969/j.issn.1674-8115.2024.01.009

Abstract

Abstract:

Objective ·To analyze the differences and classify hypertrophic cardiomyopathy (HCM) patients and healthy controls (HC) using short-axis cine cardiac magnetic resonance (CMR) images-derived radiomics features. Methods ·One hundred HCM subjects were included, and fifty HC were randomly selected at 2∶1 ratio during January 2018 to December 2021 in the Department of Cardiology, Renji Hospital, Shanghai Jiao Tong University School of Medicine. The CMR examinations were performed by experienced radiologists on these subjects. CVI 42 post-processing software was used to obtain left ventricular morphology and function measurements, including left ventricular ejection fraction (LVEF), left ventricular end-diastolic volume (LVEDV) and left ventricular end-diastolic mass (LVEDM). The 3D radiomic features of the end-diastolic myocardial region were extracted from short-axis images CMR cine. The distribution of the radiomic features in the two groups was analysed and machine learning models were constructed to classify the two groups. Results ·One hundred and seven 3D radiomic features were selected and extracted. After exclusion of highly correlated features, least absolute shrinkage and selection operator (LASSO) was used, and a 5-fold cross-validation was performed. There were still 11 characteristics with non-zero coefficients. The K-best method was used to decide the top 8 features for subsequent analysis. Among them, four features were significantly different between the two groups (all P<0.05). Support vector machine (SVM) and random forest (RF) models were constructed to discriminate the two groups. The results showed that the maximum area under the curve (AUC) for the single-feature model (first order grayscale: entropy) was 0.833 (95%CI 0.685?0.968) and the maximum accuracy for the multi-feature model was 83.3% with an AUC of 0.882 (95%CI 0.705?0.980). Conclusion ·There are significant differences in both left ventricular function and left ventricular morphology between HCM and HC. The 3D myocardial radiomic features of the two groups are also significantly different. Although single feature is able to distinguish the two groups, the combination of multi-features show better classification performance.

Key words: cardiac magnetic resonance (CMR), radiomics, hypertrophic cardiomyopathy (HCM), cine sequence, short-axis image

CLC Number:

R542.2

LIU Qiming, LU Qifan, CHAI Yezi, JIANG Meng, PU Jun. Short-axis cine cardiac magnetic resonance images-derived radiomics for hypertrophic cardiomyopathy and healthy control classification[J]. Journal of Shanghai Jiao Tong University (Medical Science), 2024, 44(1): 79-86.

Figures/Tables 7

TrendMD

References 26

1	MARON B J, MARON M S. Hypertrophic cardiomyopathy[J]. Lancet, 2013, 381(9862): 242-255.
2	WANG J, BRAVO L, ZHANG J Q, et al. Radiomics analysis derived from LGE-MRI predict sudden cardiac death in participants with hypertrophic cardiomyopathy[J]. Front Cardiovasc Med, 2021, 8: 766287.
3	GEORGIOPOULOS G, FIGLIOZZI S, PATERAS K, et al. Comparison of demographic, clinical, biochemical, and imaging findings in hypertrophic cardiomyopathy prognosis: a network meta-analysis[J]. JACC Heart Fail, 2023, 11(1): 30-41.
4	LOCKIE T, ISHIDA M, PERERA D, et al. High-resolution magnetic resonance myocardial perfusion imaging at 3.0-Tesla to detect hemodynamically significant coronary stenoses as determined by fractional flow reserve[J]. J Am Coll Cardiol, 2011, 57(1): 70-75.
5	CHIRIBIRI A, HAUTVAST G L T F, LOCKIE T, et al. Assessment of coronary artery stenosis severity and location[J]. JACC Cardiovasc Imaging, 2013, 6(5): 600-609.
6	NEISIUS U, MYERSON L, FAHMY A S, et al. Cardiovascular magnetic resonance feature tracking strain analysis for discrimination between hypertensive heart disease and hypertrophic cardiomyopathy[J]. PLoS One, 2019, 14(8): e0221061.
7	NEISIUS U, EL-REWAIDY H, NAKAMORI S, et al. Radiomic analysis of myocardial native T1 imaging discriminates between hypertensive heart disease and hypertrophic cardiomyopathy[J]. JACC Cardiovasc Imaging, 2019, 12(10): 1946-1954.
8	LIU J, ZHAO S H, YU S Q, et al. Patterns of replacement fibrosis in hypertrophic cardiomyopathy[J]. Radiology, 2022, 302(2): 298-306.
9	GERMAIN P, VARDAZARYAN A, PADOY N, et al. Classification of cardiomyopathies from MR cine images using convolutional neural network with transfer learning[J]. Diagnostics (Basel), 2021, 11(9): 1554.
10	JIANG S, ZHANG L L, WANG J, et al. Differentiating between cardiac amyloidosis and hypertrophic cardiomyopathy on non-contrast cine-magnetic resonance images using machine learning-based radiomics[J]. Front Cardiovasc Med, 2022, 9: 1001269.
11	CHENG S N, FANG M J, CUI C, et al. LGE-CMR-derived texture features reflect poor prognosis in hypertrophic cardiomyopathy patients with systolic dysfunction: preliminary results[J]. Eur Radiol, 2018, 28(11): 4615-4624.
12	AVARD E, SHIRI I, HAJIANFAR G, et al. Non-contrast cine cardiac magnetic resonance image radiomics features and machine learning algorithms for myocardial infarction detection[J]. Comput Biol Med, 2022, 141: 105145.
13	SHI R Y, WU R, AN D A L, et al. Texture analysis applied in T1 maps and extracellular volume obtained using cardiac MRI in the diagnosis of hypertrophic cardiomyopathy and hypertensive heart disease compared with normal controls[J]. Clin Radiol, 2021, 76(3): 236.e9-236.e19.
14	ANTONOPOULOS A S, BOUTSIKOU M, SIMANTIRIS S, et al. Machine learning of native T1 mapping radiomics for classification of hypertrophic cardiomyopathy phenotypes[J]. Sci Rep, 2021, 11(1): 23596.
15	FAHMY A S, ROWIN E J, ARAFATI A, et al. Radiomics and deep learning for myocardial scar screening in hypertrophic cardiomyopathy[J]. J Cardiovasc Magn Reson, 2022, 24(1): 40.
16	ZHOU D, XU J, ZHAO S H, et al. CMR publications from China of the last more than 30 years[J]. Int J Cardiovasc Imaging, 2020, 36(9): 1737-1747.
17	LIU Q M, LU Q F, CHAI Y Z, et al. Papillary-muscle-derived radiomic features for hypertrophic cardiomyopathy versus hypertensive heart disease classification[J]. Diagnostics (Basel), 2023, 13(9): 1544.
18	VAN GRIETHUYSEN J J M, FEDOROV A, PARMAR C, et al. Computational radiomics system to decode the radiographic phenotype[J]. Cancer Res, 2017, 77(21): e104-e107.
19	BIAU G. Analysis of a random forests model[EB/OL]. arXiv: 1005.0208v3(2012-05-26)[2023-03-20]. https://doi.org/10.48550/arXiv.1005.0208.
20	CORTES C, VAPNIK V. Support-vector networks[J]. Mach Lang, 1995, 20(3): 273-297.
21	PEDREGOSA F, VAROQUAUX G, GRAMFORT A, et al. Scikit-learn: machine learning in python[EB/OL]. arXiv: 1201.0490v4(2018-07-05)[2023-03-20]. https://arxiv.org/abs/1201.0490.
22	YU F, HUANG H B, YU Q H, et al. Artificial intelligence-based myocardial texture analysis in etiological differentiation of left ventricular hypertrophy[J]. Ann Transl Med, 2021, 9(2): 108.
23	GILLIES R J, KINAHAN P E, HRICAK H. Radiomics: images are more than pictures, they are data[J]. Radiology, 2016, 278(2): 563-577.
24	LAMBIN P, LEIJENAAR R T H, DEIST T M, et al. Radiomics: the bridge between medical imaging and personalized medicine[J]. Nat Rev Clin Oncol, 2017, 14(12): 749-762.
25	RAISI-ESTABRAGH Z, IZQUIERDO C, CAMPELLO V M, et al. Cardiac magnetic resonance radiomics: basic principles and clinical perspectives[J]. Eur Heart J Cardiovasc Imaging, 2020, 21(4): 349-356.
26	YING X. An overview of overfitting and its solutions[J]. J Phys: Conf Ser, 2019, 1168: 022022.

Indicator	Total (n=150)	HCM (n=100)	HC (n=50)	P value
Age/year	51 (39, 60)	52 (43, 60)	46 (36, 61)	0.408
Male/n(%)	110 (73)	71 (71)	39 (78)	0.362
Weight/kg	71.2±13.1	70.5±12.8	72.7±13.7	0.671
Height/cm	170 (162, 175)	170 (160, 173)	173 (166, 180)	0.002
BMI/(kg·m^‒2)	24.2 (22.4, 26.6)	24.5 (22.5, 26.8)	23.9 (22.3, 25.3)	0.253
BSA/m²	1.79±0.21	1.77±0.20	1.83±0.22	0.419
CMR parameter
LVEF/%	66.3±7.1	67.1±7.8	64.6±5.0	0.019
LVEDV/mL	132.4±32.3	132.3±29.0	132.7±38.4	0.035
LVEDV index/(mL·m^‒2)	72.3±31.7	75.0±15.4	71.8±15.6	0.876
LVEDM/g	114.0 (90.9, 158.4)	137.1 (104.8, 178.0)	90.4 (68.7, 103.5)	0.000
LVEDM index/(g·m^‒2)	62.7 (49.9, 88.1)	77.8 (59.3, 99.0)	47.4 (42.4, 55.1)	0.000

Indicator	Total (n=150)	HCM (n=100)	HC (n=50)	P value
Age/year	51 (39, 60)	52 (43, 60)	46 (36, 61)	0.408
Male/n(%)	110 (73)	71 (71)	39 (78)	0.362
Weight/kg	71.2±13.1	70.5±12.8	72.7±13.7	0.671
Height/cm	170 (162, 175)	170 (160, 173)	173 (166, 180)	0.002
BMI/(kg·m^‒2)	24.2 (22.4, 26.6)	24.5 (22.5, 26.8)	23.9 (22.3, 25.3)	0.253
BSA/m²	1.79±0.21	1.77±0.20	1.83±0.22	0.419
CMR parameter
LVEF/%	66.3±7.1	67.1±7.8	64.6±5.0	0.019
LVEDV/mL	132.4±32.3	132.3±29.0	132.7±38.4	0.035
LVEDV index/(mL·m^‒2)	72.3±31.7	75.0±15.4	71.8±15.6	0.876
LVEDM/g	114.0 (90.9, 158.4)	137.1 (104.8, 178.0)	90.4 (68.7, 103.5)	0.000
LVEDM index/(g·m^‒2)	62.7 (49.9, 88.1)	77.8 (59.3, 99.0)	47.4 (42.4, 55.1)	0.000

Feature class and name	HCM	HC	P value
GLDM: large dependence high gray level emphasis	34 779.60	23 528.60	0.047
First-order grayscale: kurtosis	5.92	5.05	0.116
First-order grayscale: entropy	4.03	4.38	0.004
GLSZM: zone entropy	7.76	7.67	0.037
GLCM: correlation	0.86	0.84	0.077
First-order grayscale: 90th percentile	129.70	174.60	0.054
Shape: major axis length	94.00	88.20	0.037
Shape: least axis length	75.10	72.50	0.322

Feature class and name	HCM	HC	P value
GLDM: large dependence high gray level emphasis	34 779.60	23 528.60	0.047
First-order grayscale: kurtosis	5.92	5.05	0.116
First-order grayscale: entropy	4.03	4.38	0.004
GLSZM: zone entropy	7.76	7.67	0.037
GLCM: correlation	0.86	0.84	0.077
First-order grayscale: 90th percentile	129.70	174.60	0.054
Shape: major axis length	94.00	88.20	0.037
Shape: least axis length	75.10	72.50	0.322

Feature class and name	AUC (95%CI)
Feature class and name	SVM	RF
GLDM: large dependence high gray level emphasis	0.758 (0.520‒0.863)	0.668 (0.486‒0.825)
First-order grayscale: kurtosis	0.726 (0.517‒0.866)	0.623 (0.445‒0.803)
First-order grayscale: entropy	0.833 (0.695‒0.968)	0.679 (0.435‒0.782)
GLSZM: zone entropy	0.712 (0.365‒0.801)	0.648 (0.449‒0.836)
GLCM: correlation	0.707 (0.507‒0.857)	0.667 (0.511‒0.845)
First-order grayscale: 90th percentile	0.752 (0.544‒0.910)	0.697 (0.446‒0.801)
Shape: major axis length	0.673 (0.282‒0.735)	0.554 (0.323‒0.677)
Shape: least axis length	0.680 (0.318‒0.740)	0.810 (0.625‒0.912)