Linear model for deriving the posterior ECG leads V7-V12 in healthy subjects
Author
Nagenthiraja, Kartheeban
Term
4. term
Publication year
2008
Abstract
Dette projekt udviklede og validerede en lineær regressionsmodel, der udleder de posteriore EKG-afledninger V7–V12 ud fra de otte uafhængige afledninger i det standard 12-aflednings EKG hos raske personer. I alt blev 37 raske frivillige inkluderet, og det standard EKG blev udvidet med seks posteriore afledninger placeret på begge sider af ryggen. For hver deltager blev fem på hinanden følgende 10-sekunders 18-afledningsoptagelser registreret med 500 Hz, data blev opdelt i trænings- og testgrupper, og træningsdata blev sammenlagt til ét datasæt. Modellen blev estimeret med mindste kvadraters metode, forudgået af korrelationsanalyse og bedste delmængde-regression, og regressionskoefficienterne blev tolket i lyset af et dipolhjerte med homogen ledende omgivelse. Posterior V7–V9 blev overvejende bestemt af elektrisk aktivitet i det horisontale laterale plan, mens information fra det horisontale anteriore-posteriore plan (og for V10–V12 også det vertikale plan) blev dæmpet gennem transformations- og udvælgelsestrinnene. Modellen gav acceptabel nøjagtighed hos raske, og præcisionen forbedredes yderligere ved segmentvis estimering; en pooled tilgang var mest effektiv. Samlet set er resultaterne lovende for raske, men anvendeligheden til patienter med akut myokardieinfarkt eller forstyrrede elektriske ledningsveje er usikker.
This project developed and validated a linear regression model to derive the posterior ECG leads V7–V12 from the eight independent leads of the standard 12-lead ECG in healthy subjects. Thirty-seven healthy volunteers were included, and the standard ECG was extended with six posterior leads placed on both sides of the back. For each participant, five consecutive 10-second 18-lead recordings were acquired at 500 Hz, data were split into training and test groups, and training recordings were pooled into a single dataset. The model was fitted using least squares, preceded by correlation analysis and best-subsets regression, and the regression coefficients were interpreted under a cardiac dipole assumption within a homogeneous conductive torso. Posterior leads V7–V9 were predominantly determined by electrical activity in the horizontal lateral plane, while information from the horizontal anterior–posterior plane (and for V10–V12 also the vertical plane) was attenuated by the transformation and selection steps. The model achieved acceptable accuracy in healthy subjects, with precision further improved by segment-wise estimation; a pooled method performed best. Overall, the results are promising for healthy populations, but applicability to patients with acute myocardial infarction or disturbed conduction pathways remains uncertain.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents