Transcript mutual information

國立雲林科技大學
National Yunlin University of Science and Technology
f-information measures in medical
image registration
Presenter : Ai-Chen Liao
Authors : Josien P.W. Pluim, J.B. Anotoine Maintz,
and Max A. Viergever
2004 . MI . Page : 1508 - 1516
1
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Medical Image Registration



Registration is necessary in order to be able to compare or
integrate the data obtained from different measurements.
Image registration is the process of transforming the
different sets of data into one coordinate system.
Medical image registration (e.g. for data of the same
patient taken at different points in time)
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Outline






Motivation
Objective
Method
Experiment
Conclusion
Comments
3
Intelligent Database Systems Lab
mutual information
N.Y.U.S.T.
I. M.
Motivation



In the mid 1990s, mutual information made its entrance
into the field of medical image registration.
A measure for registration of medical images that
currently draws much attention is mutual information.
Apart from mutual information, information theory
comprises many more information measures, which could
be considered for image registration.
I
-information
Whether I 1 is the optimal registration measure of I the
class of measures.
What the influence of the order
is on the registration
4
results.

Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Objective

This paper compares the performance of mutual
information as a registration measure with that of
other f -information measures.
5
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Method ─ f-divergence

The extent to which two probability distributions differ
can be expressed by a so-called measure of divergence.

A specific class of divergence measures is the set of
f  divergence measures.
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Method ─ f-information

A special case of f  divergence are the
f  information measures.
Intelligent Database Systems Lab
Method ─
N.Y.U.S.T.
I. M.
I-information
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Method
Mutual information
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Experiment
10
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Experiment
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Conclusion

In this paper we compare mutual information with several
other f -information measures by applying them to the
registration of clinical MR, PET, and CT images.

For the accuracy experiments, significantly better results
were found for several measures, compared to mutual
information.
12
Intelligent Database Systems Lab
N.Y.U.S.T.
I. M.
Comments

Advantage


Drawback


…
…
Application

The field of medical image registration
13
Intelligent Database Systems Lab