(688e) Elimination of Noise and Sensor Drifts in Pediatric Pressure Transducers for the Prediction of Central Venous Catheter-Related Thrombosis Events

Background: Central venous catheter (CVC)-related thrombosis is a common complication in the critical care setting. Vessel damage upon CVC insertion, catheter obstruction of the vessel lumen, endothelial damage from continued catheter movement within the vein, and patient specific risk factors such as sepsis, previous venous thromboembolism, renal failure and malignancy all place patients at risk for the development of catheter-related thrombosis (CRT). In the adult population, hospital acquired venous thromboembolism is one of the most common causes of preventable morbidity and mortality. Yet, there are limited data and few evidence-based guidelines for managing pediatric patients with CRT, let alone predicting/detecting when CRT has occurred. As a step toward developing predictive measures of pediatric CRT, we developed a strategy for removing baseline drift and high frequency sensor noise from pressure waveforms.

Methods: Patients in the pediatric cardiac intensive care unit at Dell Children’s Medical Center from December 2018 to January 2021 with ultrasound-confirmed CRT were included in this single center retrospective trial. Type of CVC (peripherally-inserted central catheter versus central venous line), location of the CVC, size of the CVC, and date of CVC placement, and age of patient at the time of CVC placement were all recorded. Pressure waveforms were sampled at 125 Hz and recorded with the Sickbay Platform (Medical Informatics Corp., Houston, TX). High frequency noise and low frequency baseline drift was removed with fast iterative filtering (FIF) [1].

Results: The baseline drift was successfully removed by removing IMFs 1-2 using an FIF approach over a 5 min time frame. Likewise, noise was removed by removing IMFs 6-7 and the underlying signal is reconstructed from IMFs 3-5. Preliminary evidence suggests that there are metrics in the decomposed signal (e.g., variances of the IMFs) that can be used to predict pediatric CRT occurrence.

Implications: We now have a way to analyze pressure waveform changes and produce clean signals for predictive analytics to provide early evidence of CRT in pediatric population. Further studies will extend this approach to real-time signal processing. After adding a suitable classifier to classify CRT, we will move to a prospective clinical trial.

Image caption: Demonstration of FIF-based noise and baseline drift removal from a short 9 min time segment: (A) the original signal, (B) the reconstructed signal, and (C) zoomed-in regions highlighting filter performance.

References: [1] Cicone, A. and Zhou, H., 2021. Numerical analysis for iterative filtering with new efficient implementations based on FFT. Numerische Mathematik, 147(1), pp.1-28.