Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Objective Selection of Hyperparameter for EIT Brad Graham, Andy Adler
 

Summary: Objective Selection of Hyperparameter for EIT
Brad Graham, Andy Adler
School of Information Technology and Engineering (SITE)
University of Ottawa, Canada
June 7, 2005
Introduction
Electrical impedance tomography (EIT) uses body surface electrodes to make measurements from which
an image of the conductivity distribution is calculated. Reconstructions are challenging due to the ill-
conditioning of the system matrix and susceptibility of the data to corruption from noise. Regularization
methods are used to calculate stable reconstructions by imposing additional conditions, such as image
smoothness, on a solution. A difficulty with experimental and clinical EIT reconstruction algorithms is
the requirement to select a scalar hyperparameter, , to control the amount of regularization used to
achieve a good reconstruction.
In the broader field of inverse problems work on automatic hyperparameter selection has produced
methods such as Generalized Cross Validation (GCV), L-Curve, and the Discrepancy Principle (Hansen
1992). However within the field of EIT by far the most common method of hyperparameter selection is
heuristic selection; researchers inspect reconstructions for a range of hyperparameter values and select
one.
The absence of objective hyperparameter selection methods results in several issues: 1) users of clini-
cal EIT systems would be uncomfortable using "fiddle" adjustments to modify images, 2) comparisons of

  

Source: Adler, Andy - Department of Systems and Computer Engineering, Carleton University

 

Collections: Computer Technologies and Information Sciences