2010 Design of Medical Devices Conference Abstracts

Predictive Modeling of Transplant-Related Mortality OPEN ACCESS

[+] Author and Article Information
Feng Cai

 University of Minnesota, Twin Cities

J. Med. Devices 4(2), 027527 (Aug 10, 2010) (1 page) doi:10.1115/1.3443322 History: Published August 10, 2010


This paper describes the application of machine learning approaches for predictive modeling to improve the estimation of risks for complications of allogeneic hematopoietic cell transplantation (HCT) including relapse, graft-versus-host disease, and transplant-related mortality (TRM). Clinical disease and demographic factors known to impact the outcome of HCT include: recipient and donor age, type of donor (related/unrelated), donor-recipient gender, diagnosis and disease status pre-HCT, and stem cell source (peripheral blood, marrow, and umbilical cord blood). However, biostatistical analysis of risk has only limited accuracy in estimating a given patient’s risks of serous post-HCT complications. We describe the application of standard support vector machine (SVM) classifiers for data-analytic modeling of TRM. The goal is to predict the binary output TRM (alive or dead) from a set of genetic, demographic, and clinical inputs. Classification decision rule is estimated using SVM approach appropriate for such sparse multivariate data. This study compares several feature selection techniques for modeling TRM and objectively evaluates the quality of feature selection via prediction accuracy of the corresponding SVM classifiers. In addition, we discuss methods for interpretation of multivariate SVM models.

Copyright © 2010 by American Society of Mechanical Engineers
This article is only available in the PDF format.






Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In