|
Automated Chest Health Analysis
Yiting Xie and Shuang Liu
Fully-automated chest health analysis consists of the segmentation of anatomical
regions and measuremnt of disease related biomarkers in the chest. By using a
fully-automated approach, it is possible to develop and validate computer
algorithms on very large image datasets. The key components of the automated
analysis are:
- Fully-automated segmentation algorithms that may be incrementally advanced
in ability and capability over time;
- Validation scheme that is applicable to very large datasets with tens of
thousands of CT scans;
- Documentation and updating scheme with the ability to easily incorporate
new data and algorithms.
This project is currently in development. So far the automated algorithms have
been validated on large datasets with more than 7,000 chest CT scans on eight
anatomical regions: airway, lungs, skin surface, cardiac region (aorta, heart
and pulmonary trunk), ribs, vertebrae, sternum and breast region [2].
Presentations and Publications
- A. P. Reeves. A development methodology for automated measurement of
quantitative image biomarkers: analysis of chest CT images.
OSA 2013 Imaging and Applied Optics Congress: Quantitative Medical
Imaging, Jun. 2013.
- A. P. Reeves, S. Liu, and Y. Xie. Image segmentation evaluation for
very-large data sets.
SPIE Medical Imaging, Mar. 2016.
List of Current Research Projects
|
