Trauma Care Classification. Phase 1

Abstract

This Phase I project developed a trauma care classification method based on variables that can be easily ascertained in the field environment. The major achievements of the Phase I study include: (1) Establishment of a Gaussian Potential Function Network (GPFN) architecture that allows the discrimination between various classes representing the degree of severity of the trauma classification problem. These classes constitute the basis for field triage. The GPFN is configured as an aggregate of Gaussian Potential Function Units (GPFUs); (2) Demonstration of the convergence properties of the training algorithm for the GPFN which adjusts the amplitudes, the means and the covariance matrices of the GPFUs to effect characterization of a given class as an integer value declaration; (3) Utilization of the fuzzy c-means clustering algorithm to partition the data into compact sets over which the GPFUs can be assigned. A cluster membership validity measure is also used to provide the fuzzy c-means algorithm with an estimate of the number of clusters present in the data; (4) A direct encoding classification method is also presented that allows the direct encoding of the prevalence of a given feature vector among the various classes.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 1996
Accession Number
ADB218709

Entities

People

  • Ching-fang Lin

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Biomedical Research
  • Cardiovascular Physiological Phenomena
  • Computational Science
  • Computers
  • Corporations
  • Data Analysis
  • Data Processing
  • Data Sets
  • Databases
  • Health Services
  • Information Science
  • Machine Learning
  • Neural Networks
  • Standards
  • Statistics
  • Training

Readers

  • Neural Network Machine Learning.
  • Regression Analysis.
  • Trauma or Military Medicine