Liver Segmentation Using Active Learning

1
Liver Segmentation Using Active Learning

• Ankur Bakshi
• Allison Petrosino
• Advisor Dr. Jacob Furst
• August 21, 2008

2
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

3
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

4
Introduction

• Liver has many important functions
• Liver cancer is 4th most common malignancy in the
  world
• Computed Tomography (CT) scans are a common tool
  for diagnosis

5
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

6
Problem Statement

• Liver Segmentation is an important first step for
  Computer-Aided Diagnosis (CAD)
• Difficulties associated with liver segmentation
• Time consuming
• Similarities to other organs

Source Comparison and Evaluation of Methods for
Liver Segmentation from CT datasets, Heimann et
al., 2008
7
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

8
Related Work

• Heimann et al.- statistical shape based
  segmentation
• Susomboon et al.- hybrid liver segmentation
• Tur et al.- natural language application
• Tong et al.- text classification
• Turtinen et al.- texture application
• Prasad et al.- emphysema classification

9
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

10
Liver Segmentation Algorithm
11
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

12
Methods Explored

• Passive Learning
• Active Learning
• 1000 vs 100 initial examples
• 100 vs 10 examples added
• Negatives taken from evaluated non-liver vs. all
  non-liver
• Most informative vs Hierarchical
• Gabor

13
Hierarchical Method
14
Post-Processing
15
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

16
Results, Patient 1
Method Score
Passive Learning 55
Confidence Interval 36
Active Learning, 1000 initial examples 81
Active Learning, 100 initial examples 79
Active Learning, 100 examples added 79
Active Learning, 10 examples added 55
17
Results, Patient 1
Method Scores
10 added non-evaluated 55
10 added non-liver evaluated 82
Most Informative 78
Hierarchical 77
Average Human, non-radiologist 75
18
Results, Patient 1
Slice 134
Slice 135
Slice 136
Slice 137
Slice 138
Slice 139
19
Results, Patient 3
Approach Scores
Passive 0
Confidence Interval 0
Active Learning 22
20
Results, Patient 20
Approach Scores
Passive -
Confidence Interval 59
Active Learning 50
21
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions

22
Conclusion

• Classifier based approach outperforms confidence
  interval based approach
• Active learning outperforms passive learning
• Different active learning methods have similar
  results
• 10 examples, evaluated non-liver is most
  promising
• Interesting structures highlighted for
  application in CADx systems

23
Agenda

• Introduction
• Problem Statement
• Related Work
• Liver Segmentation
• Methods
• Results
• Conclusion
• Questions