Random Forest Algorithm Construction for the Diagnosis of Coronary Heart Disease Based on Echocardiography Video Data Streams
DOI:
https://doi.org/10.20535/ibb.2021.5.1.225794Keywords:
random forest, diagnostics, speckle-tracking echocardiography, echostresstest with dobutamine, coronary heart diseaseAbstract
Background. Recent studies show that cardiovascular diseases, including coronary heart disease, are the leading causes of death and one of the main factors of disability worldwide. The detection of cases of this type of disease over the past 30 years has increased from 271 million to 523 million and the number of deaths – from 12.1 million to 18.6 million. Cardiovascular diseases are the main cause of death among the population of Ukraine and, according to this indicator, the country remains one of the world leaders. Coronary heart disease is the leading factor in the loss of health in Ukraine and modern diagnostic methods, including machine learning algorithms, are increasingly being used for timely detection.
Objective. According to the data of speckle-tracking echocardiography using the random forest method, construct classification algorithms for diagnosing violations of the kinematics of left ventricular contractions in patients with coronary heart disease at rest, and when using an echostress test with a dobutamine test.
Methods. Speckle-tracking echocardiography was used to examine 40 patients with coronary heart disease and 16 in whom no cardiac pathology was found. Echocardiography was recorded in B mode in three positions: along the long axis, in 4-chamber, and 2-chamber positions. In total, 6245 frames of the video stream were used: 1871 – without cardiac abnormalities, and 4374 – in the presence of pathology during the examination. 56 patients (2509 frames of video data) were examined without the use of a dobutamine test and 38 patients (3736 frames of video data) – using an echostress test with a dobutamine test if no disturbances were found at rest. Dobutamine doses of 10, 20, and 40 mcg were administered under the supervision of an anesthesiologist. The data of texture analysis of images were used as informative features. To build an algorithm for detecting coronary heart disease the random forest algorithm was applied.
Results. At the first stage of the study, the diagnostic algorithms norma–pathology for the state of rest and dobutamine doses of 10, 20, and 40 mcg were constructed. Before applying the algorithm the samples were randomly divided into training (70%) and test (30%). The classifiers were evaluated for accuracy, sensitivity, and specificity. According to the test samples, the accuracy of diagnostic conclusions varied from 97 to 99%. At the second stage of the study, to increase the versatility of the models, the classifier was built for all images, without dividing them into dobutamine doses. The accuracy for the test samples also ranged from 96.6 to 97.8%. To construct diagnostic algorithms by the random forest method the data of texture analysis of images were used.
Conclusions. High-precision classification models were obtained using the random forest algorithm. The developed models can be applied to the analysis of echocardiograms obtained in B mode on equipment that is not equipped with the speckle tracking technology.
References
Voigt JU, Pedrizzetti G, Lysyansky P, Marwick TH, Houle H, Baumann R, et al. Definitions for a common standard for 2D speckle tracking echocardiography: consensus document of the EACVI/ASE/Industry Task Force to standardize deformation imaging. Eur Heart J Cardiovasc Imaging. 2015;16(1):1-11. DOI: 10.1093/ehjci/jeu184
Lazoryshynets VV, Kovalenko VM, Rudenko AV, Ivaniv YA, Beshlyaga VM, Potashev SV, et al. Definition for a com-mon standard for 2D speckle-tracking echocardiography (the Association of cardiovascular surgeons of Ukraine and the Ukrainian society of cardiology working group draft consensus). Cardiol Card Surg Contin Prof Dev. 2019;2:105-29.
DOI: 10.30702/ccs.201905.02.2dst105129
Contreras-Ojeda SL, Sierra-Pardo C, Dominguez-Jimenez JA, Lopez-Bueno J, Contreras-Ortiz SH. Texture analysis of ultrasound images for pneumonia detection in pediatric patients. In: Proceedings of 22nd Symposium on Image, Signal Processing and Artificial Vision; 2019; Institute of Electrical and Electronics Engineers Inc. DOI: 10.1109/STSIVA.2019.8730238
Molinari F, Caresio C, Acharya UR, Mookiah MRK, Minetto MA. Advances in quantitative muscle ultrasonography using texture analysis of ultrasound images. Ultrasound Med Biol. 2015 Sep 1;41(9):2520-32. DOI: 10.1016/j.ultrasmedbio.2015.04.021
Kayal EB, Kandasamy D, Khare K, Bakhshi S, Sharma R, Mehndiratta A. Texture analysis for chemotherapy response evaluation in osteosarcoma using MR imaging. NMR Biomed. 2021 Feb 1;34(2):17. DOI: 10.1002/nbm.4426
Boulkenafet Z, Komulainen J, Hadid A. Face spoofing detection using colour texture analysis. IEEE Trans Inf Forensics Secur. 2016;11(8):1818-30. DOI: 10.1109/tifs.2016.2555286
Kavitha P, Prabakaran S. Designing a feature vector for statistical texture analysis of brain tumor. Int J Eng Adv Technol. 2019;8(5):1228-30.
Nastenko Ie, Dykan I, Tarasiuk B, Pavlov V, Nosovets O, Babenko V, Kruglyi V, Soloduschenko V. GMDH classification of liver states in diffuse diseases based on statistical features of the ultrasound images texture. Induktivne Modelliuvannia Skladnyh System. 2020;11:54-66.
Zayed N, Elnemr HA. Statistical analysis of haralick texture features to discriminate lung abnormalities. Int J Biomed Imaging. 2015;2015:267807. DOI: 10.1155/2015/267807
Vrbik I, Van Nest SJ, Meksiarun P, Loeppky J, Brolo A, Lum JJ, et al. Haralick texture feature analysis for quantifying radiation response heterogeneity in murine models observed using Raman spectroscopic mapping. PLoS One. 2019;14(2):e0212225. DOI: 10.1371/journal.pone.0212225
Rini Novitasari DC, Lubab A, Sawiji A, Asyhar AH. Application of feature extraction for breast cancer using one order statistic, glcm, glrlm, and gldm. Adv Sci Technol Eng Syst. 2019;4(4):115-20. DOI: 10.25046/aj040413
Ali J, Khan R, Ahmad N, Maqsood I. Random forests and decision trees. Int J Comp Sci Iss. 2012;9(5):272-8.
Sharma H, Kumar S. A Survey on decision tree algorithms of classification in data mining. Int J Sci Res. 2016 Apr 5;5(4):2094-7. DOI: 10.21275/v5i4.nov162954
Fawagreh K, Gaber MM, Elyan E. Random forests: From early developments to recent advancements. Syst Sci Control Eng. 2014;2(1):602-9. DOI: 10.1080/21642583.2014.956265
Zhong J, Liu P, Li S, Huang X, Zhang Q, Huang J, et al. A comparison of three-dimensional speckle tracking echocardiography parameters in predicting left ventricular remodeling. J Healthc Eng. 2020 Aug 1;2020:1-9. DOI: 10.1155/2020/8847144
Narula S, Shameer K, Salem OAM, Dudley JT, Sengupta PP. Machine-learning algorithms to automate morphological and functional assessments in 2D echocardiography. J Am Coll Cardiol. 2016 Nov;68(21):2287-95. DOI: 10.1016/j.jacc.2016.08.062
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2021 The autror(s)
This work is licensed under a Creative Commons Attribution 4.0 International License.
The ownership of copyright remains with the Authors.
Authors may use their own material in other publications provided that the Journal is acknowledged as the original place of publication and National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute” as the Publisher.
Authors are reminded that it is their responsibility to comply with copyright laws. It is essential to ensure that no part of the text or illustrations have appeared or are due to appear in other publications, without prior permission from the copyright holder.
IBB articles are published under Creative Commons licence:- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under CC BY 4.0 that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work.
