Post date: 08-sep-2014 14:02:42
Title: Classification of Perinatal Hypoxia
Date: 18-Feb-2014
Speaker: Ricardo Santiago Mozos (URJC) – Óscar Barquero (URJC)
Supporting material: pdf slides, Videos
Abstract: Accurate identification of Perinatal Hypoxia from visual inspection of Fetal Heart Rate (FHR) has been shown to have a limited scope. However, the use of an automated signal processing method for this purpose should be working on signals with different duration, recording interruptions for seconds and even minutes, and loose quality conditions.
We present how to classify FHR series with Normalized Compression Distance (NCD), a general methodology allowing us both the classification from the raw series, and from calculating statistical or physiological characteristics in sliding windows. We compared moments and time and frequency indices commonly used in heart rate variability analysis, in a database with 32 fetal records (17 hypoxic and 15 controls). We found NCD robust to signal loses and length differences in the series, and statistical moments were the most promising features for our data, resulting in 88% classification accuracy (92% sensitivity and 85% specificity) 3 hours before delivery.