Statistical analysis and modeling of the agreement between the Intrinsic Frequencies technique and the established cardiovascular monitoring methods
Citation & Export
Hide
Simple citation
Razavi, Marianne.
Statistical analysis and modeling of the agreement between the Intrinsic Frequencies technique and the established cardiovascular monitoring methods. Retrieved from
https://doi.org/doi:10.7282/T3GM89F7
Export
Description
TitleStatistical analysis and modeling of the agreement between the Intrinsic Frequencies technique and the established cardiovascular monitoring methods
Date Created2016
Other Date2016-05 (degree)
Extent1 online resource (xv, 185 p. : ill.)
DescriptionThere is an urgent need for a new cardiovascular monitoring technology in order to address the limitations of the traditional devices currently in use and to curb the epidemic of heart diseases. This study was designed to evaluate Intrinsic Frequencies (IFs), a novel and non-invasive cardiovascular assessment approach with the potential of rendering the monitoring process more practical and cost effective. IFs indices are extracted from the “shape” of the arterial pressure waveform via a modified sparse time-frequency method, designed for analyzing signals. Throughout this study, the performance of the IFs technique for the assessment of Left Ventricular Ejection Fraction (LVEF), Cardiac Output (CO) and Pulse Wave Velocity (PWV) was examined. The results generated by the IFs method were compared to the measurements produced by the established monitoring devices. Observational studies were conducted and through the application of supervised machine learning, numerous statistical models were produced which, displayed the relationships between the IFs technique and the traditional methods, for the evaluation of cardiovascular parameters. Multiple regression analysis was applied in order to “train” the models. The selected models were subsequently “tested” for their accuracy and precision. The limits of agreement between the IFs method and the established techniques were assessed via Bland Altman approach. There was an overall strong relationship between the IFs technique and the standard monitoring methods for the assessment of LVEF, CO and PWV. The correlation between LVEF_IFs (Model 2-iPhone) and LVEF_MRI was strong (r=0.79, p<0.0001) and Bland Altman analysis showed a reasonable clinical agreement between the two methods, with a mean bias of 1.76% and unbiased limits of agreement (LA) of +/- 17.44%. Regarding CO estimates, IFs (Model 4) was fitted on the training set only, since an appropriate testing set was unavailable at the time of the study. The results were satisfactory. The CO study revealed a significant correlation between IFs and MRI (R=0.68, p<0.0001) as well as an adequate agreement, with zero bias and narrow LA (+/- 1.78 L/min). Moreover, the generated percentage error (36%) was close to the clinically acceptable threshold (30%) for CO. In reference to PWV measurements, IFs (Model 7) displayed a moderately strong correlation with Tonometry (r=0.64, p<0.0001) and Bland Altman analysis showed a negligible bias of -0.022 m/s and LA of +/- 2.37m/s. The present study was the first to evaluate the performance of the IFs method with regards to the estimation of the major cardiovascular health indices. We demonstrated that there is a significant correlation and agreement between the IFs technique of assessing LVEF, CO, PWV and the established methods of cardiovascular monitoring.
NotePh.D.
NoteIncludes bibliographical references
Noteby Marianne Razavi
Genretheses, ETD doctoral
Languageeng
CollectionSchool of Health Professions ETD Collection
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.