30
MedIX – Summer 06 Lucia Dettori (room 745) [email protected] u

MedIX – Summer 06 Lucia Dettori (room 745) [email protected]

  • View
    221

  • Download
    4

Embed Size (px)

Citation preview

MedIX – Summer 06

Lucia Dettori (room 745)

[email protected]

Projects

Texture classification

What has been doneThings I would like to explore nextConnection to other projects

Evaluations of segmentation algorithms

Done so far …

Given a pre-segmented organ region, can you tell me what it is: kidney, heart etc?

It depends … on its texture Identify image features that give texture

information Find rules that distinguish the texture

features of one organ from another

Texture Classification Process at a glance

Apply filterTo the image

Organ/TissueSegmented Image

TextureDescriptors

Classifier(Decision Tree)

Classification rules for tissue/organs in

CT images

Step1 – Segmentation and cropping

Organs Backbone Heart Liver Kidney Spleen

Segmented 140 50 56 55 39

Cropped 363 446 506 411 364

The image might need to be cropped, when using filters that are sensitive to areas of high contrast (background)

Active Contour Mapping (Snakes) – a boundary based segmentation algorithm

Step2 – Filtering the image

Apply a filter to the image

Organ/TissueSegmented Image

For example:* Co-occurrence matrices *Run-length matrices•Wavelets•Ridgelets•Curvelets

Averages Horizontal Activity

Vertical Activity

Diagonal Activity

Wavelet transform

Haar Wavelet

8 4

9 7 3 5

6

1 -1

2 1 -1

Original image

Wavelet coefficients

6 2 1 -1

Averages Details

A D

A D

A

D

A

D

AA AD

DA DD

Step3 – Texture features extraction

Apply a filter to an image

Organ/TissueSegmented Image

TextureDescriptors

For example:Mean, standard deviation, energy, entropy etc..

Array of texture descriptors

[T1, T2, T3, …, Tn]

Step4 - Classification

Apply a filter to an image

Organ/TissueSegmented Image

TextureDescriptors

Classification rules for tissue/organs in

CT images

The process of identifying a region as part of a class (organ) based on its texture properties.

Decisiontree

Predicts the organ from the values of the texture descriptorsTraining / Testing

Classificationperformance

measures

Step5 – Evaluating the classifier

Measure Definition

Sensitivity True Positives / Total Positives

Specificity True Negatives / Total Negatives

Precision True Positives / (True Positives + False Positives)

Accuracy (True Positives + True Negatives) / Total Samples

Actual CategoryBackbone Heart Liver Kidney Spleen Total

Predicted Backbone 182 6 1 6 0 195Category Heart 3 18 4 0 0 25

Liver 0 3 30 1 7 41Kidney 10 4 0 49 1 64Spleen 0 0 4 0 8 12Total 195 31 39 56 16 337

Misclassification matrix

Performance Measures

Organ Descriptor Sensitivity Specificity Precision Accuracy

Backbone

Wavelet 82.6 96.1 82.6 93.7

Ridgelet 91.5 99.3 96.8 98.0

Curvelet 99.4 98.8 95.3 98.9

Heart

Wavelet 59.0 92.1 67.0 85.0

Ridgelet 82.5 97.5 88.5 94.6

Curvelet 89.7 99.0 95.5 97.1

Kidney

Wavelet 77.7 91.4 69.9 88.6

Ridgelet 95.4 93.3 82.0 93.8

Curvelet 96.0 98.1 93.5 97.6

Liver

Wavelet 87.3 94.4 82.6 92.8

Ridgelet 86.9 95.9 84.4 94.0

Curvelet 95.9 98.5 94.3 98.0

Spleen

Wavelet 65.5 94.3 69.7 89.5

Ridgelet 76.9 97.6 88.0 93.8

Curvelet 91.8 98.9 94.9 97.6

Average

Wavelet 74.4 93.7 74.4 89.9

Ridgelet 86.6 96.7 88.0 94.8

Curvelet 94.6 98.7 94.7 97.9

Apply a filter to an image

Organ/TissueSegmented Image

TextureDescriptors

Classification rules for tissue/organs in

CT images

Decisiontree

Things I would like to explore

Gabor filtersFractal Dimensions

Performancemeasures

Different patientsDifferent organs Abnormal texture

Different modalities

Connections to other project

Can we use wavelet, ridgelet, curvelet-based texture descriptors for content based image retrieval?

Can we use these descriptors in the volumetric segmentation?

Instead of many 2D images, can we use the same process for 3D stack of slices?

Projects

Texture classification

Evaluations of segmentation algorithms

What has been doneThings I would like to explore nextConnection to other projects

Texture segmentation Given an image, can you tell me

how many organs you have? That was easy enough. Can you

tell which organs they are?

Identifying regions with similar texture Identifying which texture it is to label the

organ

A couple of key questions

Can you do it better by varying a parameter? How do you choose the values of your segmentation parameters?

If it looks better is it really better?

A couple of key questions

Parameter optimization

Performance evaluation

1 3

4 2

0.87 0.56

0.50 0.75

Increasing value of a segmentation parameter

GroundTruth

Regionskey

Machine

Segmentations

How do I decide what the optimal value of the parameter is?

How good a segmentation is it?

The “goodness” metric

A single value that assigns a rating to a particular segmentation based on how well the machine segmented regions “match” the regions in the ground truth images

Region Categories

Ground Truth vs. Machine Segmented Correctly Detected Over Segmented Under Segmented Missed Noise

GT

MS

CORRECTLY DETECTED

OVER SEGMENTED

UNDER SEGMENTED

A Missed region is a GT region that does not participate in any instance of CD, OS, or US

A Noise region is an MS region that does not participate in any instance of CD, OS, or US

Index for each region

The “Goodness” Metric good = Correct Detection Index bad = 1-Correct Detection Index goodness = good-bad*weight

1.0

-1.0

Ceiling = CDind

Floor = 2*CDind-1

Weight Range = CDind-1

How can we use the metric?

Create a set of ground truth mosaic using radiologist-labels images of pure patches of organ tissues

Apply segmentation algorithm Optimize the segmentation parameters using the

metric Apply optimized algorithm to the “real” image

Ground Truth Region key

T=1000; GM= - .94

T=4000; GM= .74

T=3000; GM= .73T=2000; GM= - .02

T=5000; GM= .75 T=6000; GM= .08

Done so far

Used the metric on a block-wise walevet-based segmentation algorithm on some sample mosaic

To be done

Fully test the metric on a wide range of segmentation algorithms

Decouple the various components of the metric and test the individual performance measures instead of the overall score

Extend the metric to measure one region vs background segmentation

To be done

Improve the wavelet-based algorithms we have implement to include other texture features

Explore and compare other texture-based segmentation algorithm

Use regions and metric to calculate changes in time of an abnormal region

Connections to other projects

Use one of these algorithms to create a rough segmentation that will generate the starting point for a more sophisticated segmentation algorithm.

Some references ”Wavelet-based Texture Classification of Tissues in Computed

Tomography”, L. Semler, L, Dettori, and Jacob Furst.18th IEEE International Symposium on Computer-based Medical Systems, Dublin, Ireland, June 2005.

“Ridgelet-based Texture Classification in Computed Tomography”, L. Semler, L. Dettori. and W.Kerr. 8th IASTED International Conference on Signal and Image Processing, Honolulu, HW, August 2006.

“Curvelet-based Texture Classification of Tissues in Computed Tomography”, L. Semler, & L. Dettori. International Conference on Image Processing, Atlanta, GA, October 2006.

“A Comparison of Wavelet-based and Ridgelet-based texture classification of Tissues in Computed Tomography”, with Lindsay Semler, International Conference on Computer Vision Theory and Applications, Setubal, Portugal, February 2006

“A Methodology and Metric for Quantitative Analysis and Parameter Optimization of Unsupervised, Multi-Region Image Segmentation”, William Kerr, Lucia Dettori, and Lindsay Semler, 8th IASTED International Conference on Signal and Image Processing, Honolulu, HW, August 2006.