Related Experiment Videos
Lumbar disc localization and labeling with a probabilistic model on both pixel and object features.
Jason J Corso1, Raja S Alomari, Vipin Chaudhary
1Department of Computer Science and Engineering, University at Buffalo, State University of New York, USA. jcorso@cse.buffalo.edu
Medical Image Computing and Computer-Assisted Intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
|November 5, 2008
Summary
Accurate lumbar disc assessment needs precise localization and labeling. A novel two-level probabilistic model enhances this process by integrating appearance and location data for improved disc identification.
Area of Science:
- Medical imaging analysis
- Computational anatomy
Background:
- Accurate quantitative assessment of intervertebral disc pathology is crucial.
- Reliable localization and labeling of lumbar spine discs are necessary for this assessment.
Purpose of the Study:
- To develop and validate a two-level probabilistic model for accurate lumbar disc localization and labeling.
- To improve the robustness and efficiency of intervertebral disc analysis.
Main Methods:
- A two-level probabilistic model integrating pixel-level appearance and object-level relative location information.
- Generalized expectation-maximization algorithm for efficient and convergent localization and labeling.
Main Results:
- Achieved 96% accuracy in localization and labeling on 20 normal lumbar spine cases.
- Demonstrated a promising extension of the model for pathology cases.
Conclusions:
- The proposed two-level probabilistic model offers an accurate and efficient method for lumbar disc localization and labeling.
- This approach enhances robustness against ambiguous disc signatures and structural variations, paving the way for improved disc pathology assessment.
