An extended Kalman filter with inequality constraints for real-time detection of intradialytic hypotension

Intradialytic hypotension (IDH) is the most common complication of hemodialysis, affecting 15-50% of all dialysis sessions. Previously, we had presented a non-invasive Polyvinylidene Fluoride (PVDF) based sensor in the form of a ring to measure vascular tone and we showed that the morphology of the...

Full description

Saved in:
Bibliographic Details
Published in:Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Vol. 2017; pp. 2227 - 2230
Main Authors: Ansari, Sardar, Molaei, Somayeh, Oldham, Kenn, Heung, Michael, Ward, Kevin R., Najarian, Kayvan
Format: Conference Proceeding Journal Article
Language:English
Published: United States IEEE 01.07.2017
Subjects:
Adaptation models
Algorithms
Computational modeling
Covariance matrices
Electrocardiography
Electronic mail
extended Kalman filter
Humans
Hypotension
inequality constraints
Intradialytic hypotension
Kalman filters
Mathematical model
Normal Distribution
Polyvinylidene Fluoride
reflection waves
Renal Dialysis
ISSN:1557-170X, 2694-0604, 2694-0604
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Intradialytic hypotension (IDH) is the most common complication of hemodialysis, affecting 15-50% of all dialysis sessions. Previously, we had presented a non-invasive Polyvinylidene Fluoride (PVDF) based sensor in the form of a ring to measure vascular tone and we showed that the morphology of the signal can be utilized to predict IDH. This paper presents an approach for analyzing the PVDF signal using extended Kalman filter (EKF) and a synthetic model that has previously been used to model the ECG signal with Gaussian functions. Moreover, a novel approach for incorporating state inequality constraints into the EKF process using a gradient projection method is introduced. The taut string algorithm was first used to estimate the outline of the signal and remove it to highlight the reflection waves. Then, the EKF was used to characterize the morphology of the signal using Gaussian functions. The amplitudes of the Gaussian functions were used as features to train a classifier. The results indicated that the PPV and NPV for the prediction were 83.33% and 100%, respectively.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:1557-170X
2694-0604
2694-0604
DOI:10.1109/EMBC.2017.8037297