Video Segmentation - Stage I: The Critical View

Download Report

Transcript Video Segmentation - Stage I: The Critical View

Computational Image
Classification
UMBC Department of Computer Science
eBiquity Research Group
February 19, 2010
Overview




Introductions
Image Classification
Initial Results
Future Efforts
Introductions

Faculty







Yelena Yesha, PhD
Michael Grasso, MD, PhD
John Dorband, PhD
Tim Finin, PhD
Milt Halem, PhD
Anupam Joshi, PhD
Graduate students


Ronil Mokashi
Darshana Dalvi
Computational Image Classification

Categorize a raster image into a finite
set of classes.



Convert raster data into feature vectors.
Support vector machine image classifier.
Metadata to map specific classes to
biological characteristics.
Image Classification Examples




Computer assisted diagnosis of prostate
and breast cancer biopsies.
Segmentation of hysteroscopy video.
Echocardiogram analysis.
Skin cancer detection.
Biomedical Imaging: from Nano to Macro, 2007;:1284-1287
IEEE TITB, 2008 May;12(3):366-376
Proceedings 27th IEEE EMBS, 2005;:5680-5683
Conf Proc IEEE Eng Med Biol Soc, 2006;1:4775-8
Related Efforts: Video Segmentation

Laparoscopic cholecystectomy videos.


378 representative images from 5 videos.
Analyzed 49 separate image features.
Related Efforts: Video Segmentation

Image classification.



Distance metric to identify best features.
Support vector machine image classifier.
Accuracy of 91%.
Video
Segments
Frames
Related Efforts: Video Segmentation

Future directions.



Real-time analysis to assess patient safety.
Time and motion analysis of surgical
instruments.
Classification of pathology.


Hiatal hernias.
Capsule endoscopy.
Related Efforts: Cancer Screening

Skin caner screening.




Handheld iPhone image classifier.
Tool for primary care physicians.
Identify lesions in
need of dermatology
referral.
NIH and UMB
proposals pending.
Image Classification Approach
Feature Extraction
Images Organized by Class
Feature Models
Model Features
Feature Extraction
Unknown Image
Image Classifier
Image Features

Spectral features (color/tone).




Histogram (3D, color, binary, gray).
Distribution, size, width, mean, stdev.
Do not vary with translation and rotation.
Textural features (spatial distribution).



Gray-level co-occurrence matrix (GLCM).
Energy, entropy, contrast, correlation.
Independent of color distribution.
IEEE Transaction on Systems, Man, and Cybernetics. 1973 Nov; 3(6):610-621
Advanced Features - Context

Image segmentation.



Regions of interest (contextual features).
Threshold algorithms - Maximum Entropy,
Otsu Threshold, Watershed, etc.
Segmentation features of actin fibers.




Density - Area, Mean Gray Level.
Distribution - Centroid, Center of Mass.
Orientation - Angle, Elliptical Fit (wrt cell).
Order - Angle, Elliptical Fit (wrt fibers).
Advanced Features - Clustering

K-means clustering of image features.


Partitions images into clusters based on the
nearest mean, based on a first-order Markov
property.
Based on the assumption that images with
similar clinical features are more likely to be
found in the same cluster.
Model Development

Distance metrics



Manhattan distance, Jeffrey divergence.
Classification threshold.
Support vector machines



Machine learning methods.
An N-dimensional hyperplane optimally
separates images into categories.
This mapping is performed by a set of
mathematical functions, known as kernels.
Initial Results - Feature Analysis

Initial image classification experiment.


Evaluated 15 spectral and textural features.
Total of 11 images in 4 groups.



Focal adhesions images, actin stained.
1hdry, 1hwet, 24hdry, 24hwet.
Analysis.



Leave-one-out technique over all 11 images.
Manhattan distance.
Threshold of 5% (0.3% for histograms).
Initial Results - Feature Analysis
Trait+
Feature+
43
Trait17

60
Promising features.


Feature-
13
56
147
164
160
220




Sensitivity = 76.8%
Specificity = 89.6%
Accuracy = 86.4%

Gray-scale distribution
Medium
Mode
Homogeneity
Energy
Entropy
Inverse difference
moment
Initial Results - Segmentation

Second image classification experiment.

Evaluating 4 new image features.



Density, Distribution, Orientation, Order
Experimenting with threshold algorithms to
optimize image segmentation.
Total of 11 images in 4 groups.


Focal adhesions images, actin stained.
1hdry, 1hwet, 24hdry, 24hwet.
Initial Results - Segmentation

Orientation feature using elliptical fit.




Image moment-preserving threshold.
Elliptical fit.
Cell angle = 132°
Average (weighted) actin fiber angle = 129°.
Initial Results - Segmentation
Initial Results - Analysis Platform

To automate and optimize image
processing algorithms.
Future Efforts





Use feature analysis to develop a
support vector machine image classifier.
Continue image segmentation work.
Correlate actin data to other images.
Incorporate successful algorithms in the
Analysis Platform.
Identify ontologies to map specific
classes to biological characteristics.