Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Evaluation of partial classification algorithms using ROC curves

G Tusch1

  • 1Clinic for Abdominal and Transplantation Surgery, Medical School Hannover, 30623 Hannover, Germany.

Medinfo. MEDINFO
|January 1, 1995
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Classification of thoracic and lumbar injuries: An analysis of interobserver reliability.

Der Orthopade·2017
Same author

Evaluation of the mobility of adjacent segments after posterior thoracolumbar fixation: a biomechanical study.

European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society·2001
Same author

Replacement of calcineurin inhibitors with mycophenolate mofetil in liver-transplant patients with renal dysfunction: a randomised controlled study.

Lancet (London, England)·2001
Same author

[Surgical therapy of cystic echinococcosis of the liver].

Zentralblatt fur Chirurgie·2001
Same author

A model for optimal sequential decisions applied to liver transplantation.

Studies in health technology and informatics·2001
Same author

Data warehouse and data mining in a surgical clinic.

Studies in health technology and informatics·2001
Same journal

CHROMASSI: a therapy advice system based on chrono-massage and acupression using the method of ZiWuLiuZhu.

Medinfo. MEDINFO·1995
Same journal

"Auctoritas" psychiatric expert system shell.

Medinfo. MEDINFO·1995
Same journal

Electronic-Brain Medical Erudite: its past, present, and future.

Medinfo. MEDINFO·1995
Same journal

A case against logic.

Medinfo. MEDINFO·1995
Same journal

A simplified judgment system for medical informatics.

Medinfo. MEDINFO·1995
Same journal

A knowledge-based care protocol system for ICU.

Medinfo. MEDINFO·1995
See all related articles

This study introduces new methods for evaluating partial classification algorithms in clinical decision support. It extends traditional measures to compare diverse AI and fuzzy set models, improving diagnostic accuracy.

Area of Science:

  • Medical Decision Theory
  • Artificial Intelligence in Medicine
  • Clinical Informatics

Background:

  • Classical classification algorithms select one class, with established measures like sensitivity and error rate.
  • Partial classification, common in AI and fuzzy set theory, assigns observations to multiple classes, where traditional measures are insufficient.
  • Existing discriminant measures are not applicable to partial classification, necessitating new assessment methods for clinical decision support.

Purpose of the Study:

  • To establish a methodological framework for assessing and comparing partial classification algorithms in clinical settings.
  • To adapt and extend traditional discriminant measures for use with partial classification models.
  • To demonstrate the utility of the proposed methods using a clinical example from cranial computed tomography.

Related Experiment Videos

Main Methods:

  • Developed a framework based on medical decision theory for evaluating partial classification.
  • Extended Receiver Operating Characteristic (ROC) analysis to accommodate partial classification, requiring two cutoff points instead of one.
  • Proposed new measures analogous to sensitivity and error rate for partial classification, applicable to parametric and non-parametric models.

Main Results:

  • Demonstrated that classical ROC analysis is inappropriate for partial classification.
  • Showcased an extended ROC approach suitable for partial classification, requiring rank order of alternatives.
  • Successfully compared linear discriminant analysis, Bayesian, and fuzzy procedures in a cranial CT scan example using the new methodology.

Conclusions:

  • The proposed methodology provides a sound mathematical basis for assessing partial classification algorithms in medicine.
  • The extended ROC analysis and new discriminant measures enable comparison of diverse algorithms, regardless of their underlying approach.
  • This framework enhances the reliability of computer-aided decision support systems in clinical practice.