Using Machine Learning to Fill in Missing Values in Pulsative Data from Diverse Clinical Datasets
Author(s)
Kim, Nabin
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The integration of machine learning techniques with healthcare data has spurred innovative solutions for clinical diagnosis and patient care. This study addresses the critical challenge of missing data in physiological signals, particularly Electrocardiogram (ECG) and Photoplethysmography (PPG), which are vital for diagnosing conditions like atrial fibrillation and monitoring patient health. While existing research predominantly relies on the MIMIC-III dataset, which offers rich but limited patient diversity, this study leverages the MODS dataset from Emory University, providing a more comprehensive representation of patient demographics and clinical conditions. The primary objective is to test the generability of the performance of the BDC Transformer, a machine-learning model for imputing missing values in waveform data, on the MODS dataset. This project underscores the significance of interdisciplinary collaboration in advancing healthcare informatics, with implications for improved patient outcomes and personalized care delivery.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis