Fatty Liver Characterization and Classification by Ultrasound - PowerPoint PPT Presentation

Loading...

PPT – Fatty Liver Characterization and Classification by Ultrasound PowerPoint presentation | free to download - id: 58a619-NzdjZ



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Fatty Liver Characterization and Classification by Ultrasound

Description:

Fatty Liver Characterization and Classification by Ultrasound Ricardo Ribeiro1,2 and Jo o Sanches1,3 1Institute for Systems and Robotics 2Escola Superior de ... – PowerPoint PPT presentation

Number of Views:96
Avg rating:3.0/5.0
Slides: 15
Provided by: Jo207
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Fatty Liver Characterization and Classification by Ultrasound


1
Fatty Liver Characterization and Classification
by Ultrasound
  • Ricardo Ribeiro1,2 and João Sanches1,3
  • 1Institute for Systems and Robotics
  • 2Escola Superior de Tecnologias de Saúde de
    Lisboa
  • 3Instituto Superior Técnico

2
Short Story
  • Steatosis is mainly a textural abnormality of
    the hepatic parenchyma due to fat accumulation on
    the hepatic vesicles (genetic, alcohol and
    obesity)
  • The ultimate goal is to accurately quantity the
    degree and severity of the disease
  • US signal processing
  • Classification
  • Results

3
Motivation
  • Detection and Classification of Liver Steatosis
    by ultrasound

Normal
fatty
4
Steatosis characterization
  1. Echogenicity
  2. Texture
  3. Depth attenuation
  4. Anatomical details

5
US Image Processing
6
Speckle model
  • RF signal, y, non compressed, is Rayleigh
    distributed
  • Observation model where
  • is a unit parameter Rayleigh distribution,
    independent of f, like in the AWGN model,
    where ? is independent of f. Let us
    call it, Multiplicative White Rayleigh Noise
    (MWRN)

7
Synthetic Data
8
Real Data - Liver
Original
RF
Noiseless
Speckle
9
Intensity Features
  • Intensity decay with depth (extracted form F)
  • where

10
Textural Features
  • Energies of the first Haar wavelet decomposition
    vertical and horizontal detail fields, Ev and Eh

11
Classification - Data
  • Two sets of 10 images each of the hepatic
    parenchyma from 5 healthy subjects (w1) and from
    5 subjects with fatty liver (w2) are used.
  • The classification was made manually by the
    operators and confirmed by indicators obtained
    from laboratorial analysis.
  • Mean (??) and covariance (??) matrices for each
    class were estimated using 10 images.
  • A leaveoneout crossvalidation method is used
    to assess the performance of the classifier.

12
Classification - Feature Space
13
Classification
  • Bayes classifier,
  • where and

14
Conclusions
  • Steatosis diagnosis from 2D B-mode US images,
  • Ultrasound image decomposition noiseless and
    speckle fields,
  • Features Intensity (depth decay) and textural
    (Wavelet decomposition details),
  • Bayes classifier
  • Future
  • Steatosis quantification
  • Other features (laboratorial analysis)
  • Other classifier (SVM)
About PowerShow.com