Quantification of Myocardial Blood Flow by Machine Learning Analysis of Modified Dual Bolus MRI Examination
- PDF / 1,444,683 Bytes
- 10 Pages / 593.972 x 792 pts Page_size
- 79 Downloads / 181 Views
Annals of Biomedical Engineering (Ó 2020) https://doi.org/10.1007/s10439-020-02591-0
Original Article
Quantification of Myocardial Blood Flow by Machine Learning Analysis of Modified Dual Bolus MRI Examination MINNA HUSSO ,1 ISAAC O. AFARA,2,3 MIKKO J. NISSI,2 ANTTI KUIVANEN,4 PAAVO HALONEN,4 MIIKKA TARKIA,5 JARMO TEUHO,5 VIRVA SAUNAVAARA,5,6 PAULI VAINIO,1 PETRI SIPOLA,1 HANNU MANNINEN,1 SEPPO YLA¨-HERTTUALA,4,7 JUHANI KNUUTI,5 and JUHA TO¨YRA¨S1,2,3 1
Diagnostic Imaging Center, Kuopio University Hospital, PO Box 100, 70029 KYS Kuopio, Finland; 2Department of Applied Physics, University of Eastern Finland, Kuopio, Finland; 3School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane, Australia; 4A.I. Virtanen Institute for Molecule Sciences, University of Eastern Finland, Kuopio, Finland; 5Turku PET Centre, University Hospital and University of Turku, Turku, Finland; 6Department of Medical Physics, Turku University Hospital, Turku, Finland; and 7Heart Center and Gene Therapy Unit, Kuopio University Hospital, Kuopio, Finland (Received 31 March 2020; accepted 11 August 2020) Associate Editor Umberto Morbiducci oversaw the review of this article.
Abstract—Contrast-enhanced magnetic resonance imaging (MRI) is a promising method for estimating myocardial blood flow (MBF). However, it is often affected by noise from imaging artefacts, such as dark rim artefact obscuring relevant features. Machine learning enables extracting important features from such noisy data and is increasingly applied in areas where traditional approaches are limited. In this study, we investigate the capacity of machine learning, particularly support vector machines (SVM) and random forests (RF), for estimating MBF from tissue impulse response signal in an animal model. Domestic pigs (n = 5) were subjected to contrast enhanced first pass MRI (MRIFP) and the impulse response at different regions of the myocardium (n = 24/pig) were evaluated at rest (n = 120) and stress (n = 96). Reference MBF was then measured using positron emission tomography (PET). Since the impulse response may include artefacts, classification models based on SVM and RF were developed to discriminate noisy signal. In addition, regression models based on SVM, RF and linear regression (for comparison) were developed for estimating MBF from the impulse response at rest and stress. The classification and regression models were trained on data from 4 pigs (n = 168) and tested on 1 pig (n = 48). Models based on SVM and RF outperformed linear regression, with higher correlation (R2SVM = 0.81, R2RF = 0.74, R2linear_regression = 0.60; qSVM = 0.76, qRF = 0.76, qlinear_regression = 0.71) and lower error (RMSESVM = 0.67 mL/g/min, RMSERF = 0.77 mL/g/ min, RMSElinear_regression = 0.96 mL/g/min) for predicting
Address correspondence to Minna Husso, Diagnostic Imaging Center, Kuopio University Hospital, PO Box 100, 70029 KYS Kuopio, Finland. Electronic mail: minna.husso@kuh.fi
MBF from MRI impulse response signal. Classifier based on SVM was optimal for
Data Loading...
