Automated quality analysis and decision support for rural midwifery at the edge
Author(s)
Aghomo, Raisa-Claire Ngeniform
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
This thesis examines the optimization of exponential smoothing parameters (alpha values) for amplitude detection algorithms in fetal Doppler ultrasound quality assessment. Through systematic evaluation of seven alpha values (0.1-0.99), the research identified a fundamental trade-off between classification accuracy and notification stability. While higher alpha values achieved superior accuracy (up to 96.6% with α=0.99), they also generated significantly more frequent notification changes (95.5 changes/file), potentially affecting usability. Conversely, lower alpha values provided more stable notifications at the cost of reduced accuracy. The study introduced the "eye closing interval" as a metric for quantifying notification stability and demonstrated the potential value of a dual-threshold approach using different alpha values for low volume and high saturation detection. These findings provide a methodological framework for optimizing signal processing algorithms in resource-constrained healthcare settings, ultimately supporting improved fetal monitoring accessibility in low- and middle-income countries.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis