tissue characterization by image analysis for diagnosis
TRANSCRIPT
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Tissue Characterization by Image Analysis for Diagnosis Purposes
Biomedical signal and image processing Master in Medical Informatics
FCUP 2012
J. Miguel Sanches (PhD) Assistant Professor
Institute for Systems and Robotics Bioengineering Department (DBE)
Instituto Superior Técnico / Technical University of Lisbon
email: [email protected] home: www.isr.ist.utl.pt/~jmrs
1
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Medical Imaging
Different Biomedical Image Modalities are corrupted and distorted by different types of noise
CT MRI LSFCM
(IV)US
MRI/ASL
2
Noise
• Is usually discarded because it is unwanted and purposeless • Is generated during the acquisition process and/or processing • Is stochastic and is usually described from a statistical point of view, e.g., first or
higher order statistics • Is difficult to eliminate • It is usually very difficult to decide what is noise/artifact what what is not
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
But it can contains useful information about the observed
object
3
Image Distortion and Noise Observation model
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
)(xfy =
x
Original
Hxw =
Blurred
)1log( += wz
Non linear distortion
η+= zy
Noise
Typical effects/models
4
(Usual) Goal
Estimate x from y
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
))(())((ˆ 1 xygxyfx == −
x̂
Estimated
y
Noisy and distorted observation
x
Original
5
Ill-posedness
• Estimating x from y is usually an ill-posed inverse problem • A problem is ill-posed if it is not well-posed. • The problem is well-posed, according Hadamard, if
– A solution exists – The solution is unique – The solution depends continuously on the data
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
)(ˆ ygx =
6
Example
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
),(minargˆ xyExx
=
),( yxE
x
),( 1yxE
),( 2yxE
x1x
2x
7
Mathematical Characterization
• Noise Characterization: – Additive/Multiplicative/Other – White / Colored
e.g. White and Gaussian
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
{ }iyYXYpY =≈ )),(,( θ
),())(|(
))(|())(|(
2σθ
θθ
iiii
iiii
xNxyp
xypxYp
≈
=∏
8
White:
Gaussian:
AWGN
Additive White Gaussian Noise
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
),())(|(
),0()(
2
2
σθ
ση
η
iiii
i
iii
xNxyp
Np
xy
≈
≈
+=
9
MWGN
Multiplicative White Gaussian Noise
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
),0())(|(
),0()(
22
2
σθ
ση
η
iiii
i
iii
xNxyp
Np
xy
≈
≈
=
10
MWRN
Multiplicative White Rayleigh Noise
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
)())(|(
),(
)1,()(
2
2
iiii
xy
ii
iii
xRxyp
exyxyR
Rp
xy
≈
=
=
=
−
θ
ηη
η
11
Multiplicative in Wide Sense
• Variance depending on Signal • Not necessarily multiplicative
in the algebraic sense
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
xy
eyxxyPoissonxyp −≈≈
!),())(|( θ
The Variance depends of the observation (signal); it is not constant
12
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Speckle
Processed by José Seabra (Biomedical PhD student)
• Ultrasound images usually present low quality (low SNR) • Images are corrupted by speckle noise (multiplicative)
13
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Medical Information
Speckle Pattern (texture) contains relevant medical information
fatty Liver.
14
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Image Decomposition
• Decompose the image in textural and anatomical/morphological components.
• Characterize the texture/“noise” (speckle).
• Associate the texture characteristics with the disease.
• Classification and Quantification for Detection (Diagnosis) and Quantification (Severity Assessment) of the disease.
15
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Decomposition Examples
16
Texture Characeterization Rayleigh Mixture Model(RMM)
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012 17
Seabra, J.C.; Ciompi, F.; Pujol, O.; Mauri, J.; Radeva, P.; Sanches, J. , "Rayleigh Mixture Model for Plaque Characterization in Intravascular Ultrasound," Biomedical Engineering, IEEE Transactions on , vol.58, no.5, pp.1314-1324, May 2011.
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Speckle Analysis for Diagnostic Purposes
• Liver Steatosis – Ricardo Ribeiro (PhD Student)
• Classification and Staging of Chronic Liver Disease from Multimodal Data
– Ricardo Ribeiro (PhD Student)
• Carotid Atherosclerotic Plaques
– José Seabra (PhD) – David Afonso (PhD Student) – Manya Afonso (PhD, Post-Doc)
18
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Liver (Steatosis and Cirrhosis)
• Diffuse Liver diseases such as Steatosis and Cirrhosis are mainly textural abnormalities of the hepatic parenchyma
• Today, the assessment is subjectively performed by visual inspection
19
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Steatosis
• Comparison with histological data
20
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Liver Steatosis Local Analysis
21
Classification and Staging of Chronic Liver Disease from Multimodal Data
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Normal chronic hepatitis
Cirrhosis compensated decompensated
22
Clinical Based Classifier (CBC)
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012 23
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Atherosclerotic Plaques Carotid
24
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Tissue Characterization from Intra Vascular US (IVUS)
Collaboration with the Centre de Visió per Computador, Universitat Autònoma de Barcelona, Profª Petia Radeva
25
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Plaque IVUS
Decomposition
Classification
26
Plaque B-Mode
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012 27
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
3D Generalization US Acquisition
• Freehand • Mechanical
28
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
3D Diagnosis
Global and local analysis
29
Enhanced Activity Index (EAI) Architecture
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012 30
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
Enhanced Activity Index (EAI) Prototype
31
Computer Aided Diagnosis AtherosRiskTM
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012 32
Biomedical signal and image processing, Master in Medical Informatics, FCUP, 2012
THANKS
33