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

Ideal observer approximation using Bayesian classification neural networks.

M A Kupinski1, D C Edwards, M L Giger

  • 1Department of Radiology, University of Chicago, IL 60637, USA. kupinski@radiology.arizona.edu

IEEE Transactions on Medical Imaging
|October 5, 2001
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

SU-E-E-03: Ethics and Professionalism Education in Medical Physics: A Needs Assessment Study.

Medical physics·2017
Same author

ADAPTIVE SMALL-ANIMAL SPECT/CT.

IEEE Nuclear Science Symposium conference record. Nuclear Science Symposium·2015
Same author

Automated detection of mass lesions in dedicated breast CT: a preliminary study.

Medical physics·2012
Same author

Radiographic texture analysis of densitometer-generated calcaneus images differentiates postmenopausal women with and without fractures.

Osteoporosis international : a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA·2006
Same author

Computerized mass detection for digital breast tomosynthesis directly from the projection images.

Medical physics·2006
Same author

Computerized detection of mass lesions in digital breast tomosynthesis images using two- and three dimensional radial gradient index segmentation.

Technology in cancer research & treatment·2004
Same journal

LLM-enhanced Neuron Segmentation and Reconstruction in Complex Mouse Brain Images.

IEEE transactions on medical imaging·2026
Same journal

Matrixed-Spectrum Decomposition Accelerated Linear Boltzmann Transport Equation Solver for Fast Scatter Correction in Multi-Spectral CT.

IEEE transactions on medical imaging·2026
Same journal

The Ritz Adjoint Method for MRI Pulse Design.

IEEE transactions on medical imaging·2026
Same journal

Physiology-guided Self-supervised Learning for Simultaneous Dual-Tracer PET Separation.

IEEE transactions on medical imaging·2026
Same journal

Informed-Exploration Reinforcement Learning for Automated Virtual Coronary Intervention Planning.

IEEE transactions on medical imaging·2026
Same journal

4D Reconstruction of Fetal Left Ventricle from Echocardiography via 2.5D Radial Segmentation and Graph-Fourier Reconstruction.

IEEE transactions on medical imaging·2026
See all related articles

Bayesian artificial neural networks (ANNs) approximate ideal observers for classification tasks. With sufficient data, ANNs accurately model optimal decision variables, even with excess hidden units, but optimal performance depends on data complexity.

Area of Science:

  • Machine Learning
  • Statistical Pattern Recognition
  • Computational Neuroscience

Background:

  • Optimal classification relies on the likelihood ratio or its monotonic transformations, defining an "ideal observer."
  • Artificial neural networks (ANNs) approximate ideal observers with large datasets but can overfit with smaller samples.
  • Bayesian methods regularize ANNs, aiming to improve classifier robustness and ideal observer approximation.

Purpose of the Study:

  • To evaluate the accuracy of Bayesian ANNs in modeling ideal observer decision variables.
  • To investigate the impact of hidden unit count, signal-to-noise ratio, and data dimensionality on Bayesian ANN performance.

Main Methods:

  • Trained Bayesian artificial neural network (ANN) models.
  • Assessed model accuracy across varying numbers of hidden units, signal-to-noise ratios, and data dimensions.

Related Experiment Videos

  • Compared Bayesian ANN performance against ideal observer benchmarks.

    • Main Results:

    • Bayesian ANNs accurately approximate ideal observer decision variables when sufficient training data is available.
    • Excess hidden units in Bayesian ANNs did not significantly reduce accuracy with adequate data.
    • The optimal number of hidden units required for accurate modeling varied with data complexity.

    Conclusions:

    • Bayesian ANNs are effective tools for approximating ideal observers in classification.
    • Model performance is robust to over-parameterization (excess hidden units) given sufficient data.
    • Careful consideration of network architecture (hidden units) is necessary based on data characteristics.