Beruflich Dokumente
Kultur Dokumente
Lecture # 1
Estimation Theory
Norbert Schuff
Professor of Radiology
VA Medical Center and UCSF
Norbert.schuff@ucsf.edu
UCSF VA
Department of
Radiology & Biomedical Imaging
What Is Medical Imaging Informatics?
• Signal Processing
– Digital Image Acquisition
– Image Processing and Enhancement
• Data Mining
– Computational anatomy
– Statistics
– Databases
– Data-mining
– Workflow and Process Modeling and Simulation
• Data Management
– Picture Archiving and Communication System (PACS)
– Imaging Informatics for the Enterprise
– Image-Enabled Electronic Medical Records
– Radiology Information Systems (RIS) and Hospital Information Systems (HIS)
– Quality Assurance
– Archive Integrity and Security
• Data Visualization
– Image Data Compression
– 3D, Visualization and Multi-media
– DICOM, HL7 and other Standards
• Teleradiology
– Imaging Vocabularies and Ontologies
– Transforming the Radiological Interpretation Process (TRIP)[2]
– Computer-Aided Detection and Diagnosis (CAD).
– Radiology Informatics Education
• Etc.
UCSF VA
Department of
Radiology & Biomedical Imaging
What Is The Focus Of This Course?
Learn how to maximize information using efficient
computational tools
Generative
statistic
Collect
data Compare
with
Inference model
Statistic
UCSF VA
Department of
Radiology & Biomedical Imaging
Challenge: Maximizing Information Gain
1.0
0.8
0.6
0.4
0.2
0.0
Intensity
UCSF VA
Department of
Radiology & Biomedical Imaging
Motivation Example III:
Signal Decomposition
Diffusion Tensor Imaging (DTI) Goal:
•Sensitive to random motion of water Represent fiber bundles
•Senses the neighborhood on microscopic scale
Microscopic
tissue sample
: unknown world
state of interest
: measurement
World ˆ : Estimator - a
Measurement good guess of
based on
measurements
Cartoon adapted from: Rajesh P. N. Rao, Bruno A. Olshausen Probabilistic Models of the Brain.
MIT Press 2002.
UCSF VA
Department of
Radiology & Biomedical Imaging
Deterministic Model
N = number of measurements
M = number of states, M=1 is possible
Usually N > M and ||noise||2 > 0
Note:
1) we make no assumption about the
distribution of θ
1) Each value is as likely as any
another value
φ Hθ noise
φ Hθˆ 0
LSE
θLSE H H HT
T 1
But
•LSE makes no assumptions about distributions of data or parameters
•Has no basis for statistics “deterministic model” UCSF VA
Department of
Radiology & Biomedical Imaging
Prominent Examples of LSE
N
150 1
Mean Value: ˆmean j
100 N j 1
Intensities (Y)
N
1
50
ˆvariance j ˆmean
2
Variance:
0
N 1 j 1
-50
Measurements (x)
200
Amplitude: ˆ1
100
Frequency: ˆ2
Intensity
0 Phase: ˆ3
-100
Decay: ˆ4
100 300 500 700 900
Measurements
UCSF VA
Department of
Radiology & Biomedical Imaging
Likelihood Model
Likelihood of
We believe is governed by chance, but we
don’t know the governing probability.
Note:
is random and unknown
has been measured
The likelihood of parameter is the
probability of the observed data as a
function of this parameter
UCSF VA
Department of
Radiology & Biomedical Imaging
Likelihood Function
Let be a random variable with a discrete probability distribution p depending
on the parameter . The likelihood function of (given the outcome j of ) is:
Lθ | Pr j ; θ
Let the probability that a coin lands heads up when tossed be pH.
The probability of getting two heads in two tosses (HH) is pH * pH
Thus, if pH=0.5, the probability of seeing to heads is 0.25.
Another way of saying this is that the likelihood that pH=0.5 given
observation HH = 0.25 is
UCSF VA
Department of
Radiology & Biomedical Imaging
Likelihood Model (cont’d)
New Goal:
Find an estimator
which gives the most likely
probability of
underlying L | .
UCSF VA
Department of
Radiology & Biomedical Imaging
Maximum Likelihood Estimator (MLE)
Goal: Find estimator which gives the most likely value of
underlying the likelihood function of
Highest probability
θˆ MLE max Pr ; θ
The MLE can be found by taking the derivative of the likelihood function
UCSF VA
Department of
Radiology & Biomedical Imaging
Example I: MLE Of Normal Distribution
Normal Distribution
Normal distribution 1.0
1 2
Pr N | , exp 2 j
N
2 0.8
2 j 1 0.6
a2
0.4
ln Pr N | , 2
j
N
2 1 2 0.0
2 j 1
100 300 500
a1
700 900
1 N
d
d
ln Pr
4ˆ N j 1
2 j 0 -5
-10
N
1
MLE
N
j
j 1
-15
100 300 500 700 900
a1
UCSF VA
Department of
Radiology & Biomedical Imaging
Example II: MLE Of Binominal Distribution
(Coin Tosses)
n= # of tosses
= # of heads in n tosses
= probability of obtaining a head in a toss (unknown)
Pr(; ; n) = probability of heads from n tosses given
Pr(; =0.7;n)
0.7
0.2
0.1
y
0.2
0.1
0.0
1 2 3 4 5 6 7 8 9 10
UCSF VA
N Department of
Radiology & Biomedical Imaging
MLE Of Coin Toss (cont’d)
Goal:
Given Pr(; ), estimate the most likely probability distribution ˆMLE max Pr;
that produced the data.
# heads
Intuitively: ˆMLE
# tosses
y ! n y !
0.25
0.20
Likelihood
0.15
Likelihood function of coin tosses
0.10
n!
L | y, n y 1
n y 0.05
y ! n y ! 0.00
0.1 0.3 0.5 0.7 0.9
W
What is the likelihood of observing 7 heads
given that we tossed a coin 10 times?
10!
L | n 10, y 7 0.57 1 0.5
10 7
0.12
For a fair coin =0.5: 7!10 7 !
10!
L | n 10, y 7 0.67 1 0.6
10 7
For an unfair coin =0.6 0.21
7!10 7 !
n! 0.20
L | y y 1
n y
Likelihood
y ! n y !
0.15
0.10
0.05
0.00
W
log likelihood function
ln L | y
-1
n!
ln y ln n y ln 1
y ! n y ! -2
-3
d ln L y n y
0
d 1
y n y # heads
0 ˆMLE
1 n # tosses
According to the MLE principle, the distribution Pr(y; MLE; n) for a given n
is the most likely distribution to have generated the observed data
of y.
prior Pr
New Goal:
Find the estimator which
gives the most likely
probability distribution of
given everything we know.
Posterior
Pr | Pr ; Pr
Medical Imaging Informatics 2009, Nschuff
UCSF VA
Course # 170.03 Department of
Slide 25/31 Radiology & Biomedical Imaging
Thomas Bayes (1701-1761)
ˆMAP maxPr ; Pr
MAP can be found by taken the partial derivative of the joint density
With respect to
2 N
ˆMAP
T
2 2
j 1
j
MAP is a linear combination between the prior mean and sample mean weighted
by there respective covariances
(|)
MSE
X MAP
UCSF VA
Department of
Radiology & Biomedical Imaging
Some Desirable Properties of Estimators I:
Unbiased: Mean value of the error should be zero
E - 0
Cik E i -i k - k
T
some.very.small.value
1 N 1
Mean: E ˆ E j N
N j 1 N
The sample mean is an unbiased estimator of the true mean
1 N 2
2 E j 2 N
1
E ˆ
2 2 2
Variance:
N j 1 N N
1.0
0.8
0.6
0.4
0.2
0.0
Intensity
J.R.Jiménez-Alaniz, et al.
IEEE TRANSACTIONS ON
MEDICAL IMAGING, VOL. 25, NO. 1,
JANUARY 2006
UCSF VA
Department of
Radiology & Biomedical Imaging
Estimation Theory: Motivation Example III
Diffusion Spectrum Imaging – Human Cingulum Bundle
Goal:
Capture directions
of fiber bundles
NeuroImage 37 (2007)
116–129
UCSF VA
Department of
Radiology & Biomedical Imaging
MAP Estimation In Image Reconstruction
From: A. Greenspan in
The Computer Journal Advance Access published February 19, 2008
zDFT =
zero-filled DFT