Please use this identifier to cite or link to this item:
Title: A data-manifold based scoring system and its application to cardiac arrest prediction
Authors: Liu, Tianchi.
Keywords: DRNTU::Engineering
Issue Date: 2013
Abstract: Scoring systems have been widely used in medical applications, for example to assess the severity of illness in intensive care units (ICU). In this Final Year Project (FYP), the author researched into different ways of developing novel scoring systems for predicting cardiac arrest within 72 hours. This report is the documentation of the works done and the presentation of the novel scoring system developed. Several approaches were looked into and eventually one scoring system with best performance is proposed in this report. The proposed scoring system computes the scores based on the data manifolds possessed by training and testing data, and therefore global consistency of the data is utilized. This proposed scoring system is essentially a supervised learning process. Other approaches include semi-supervised learning and supervised learning with pre-known scores. The validation experiment is conducted on real patients’ data, including both vital signs and heart rate variability (HRV) parameters. Performance is evaluated under leave-one-out cross-validation (LOOCV) framework. Moreover, comparison of the proposed scoring system with previous work can be found in terms of sensitivity, specificity and positive predictive value (PPV), negative predictive value (NPV) and receiver operating characteristic (ROC). The proposed Data-manifold based Scoring System is able to achieve better performance in generating meaningful risk scores than the Distance-based Scoring System previously developed.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
1.12 MBAdobe PDFView/Open

Page view(s) 50

checked on Oct 21, 2020

Download(s) 50

checked on Oct 21, 2020

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.