Automatic Segmentation and Volume Estimation of Lung Cancer Nodules using Computed Tomography Timothy R. Tuinstra, PhD Candidate
The Problem: Lung Cancer is the Leading Cause of Cancer Death in the U.S. Accurate Diagnosis of Lung Cancer is possible given Accurate Volume Estimates of Pulmonary Nodules. Electrical Engineers are studying ways to estimate Nodule Volumes from Computed Tomography Images.
What is Computed Tomography? X-Ray Computed Tomography allows us to view slices of the Human Body. Image Courtesy of Siemens
Example CT Image of Lungs with Cancer Nodule Cancer Nodule (Small Tumor) Lung Wall
Diagnostic Process Identify Nodules Segment Nodules (Determine their Boundaries) Estimate Volume Track Volumes over time to determine tumor growth.
Example Automatic Nodule Segmentations
Volume Estimation Techniques: Area Method Area of largest slice Example Segmentation
Volume Estimation Techniques: Minor Axis Method Major Axis Length Minor Axis Length Example Segmentation
Volume Estimation Techniques: Perimeter Method Number of Segmented Voxels Voxel Dimensions in the x & y directions CT slice spacing s Voxel
New Technique: Minimax (developed by Dr. Russell Hardie) Number of columns in Segmentation Estimate of Column Height This technique seeks to mitigate partial volume effects.
New Techniques: Modified Winer- Muram Observation Model: Magnification: Observed Volume Magnification True Volume Slice Thickness Slice Spacing Constant
1.25 mm Slice Spacings
5 mm Slice Spacing
8 mm Slice Spacing
10 mm Slice Spacing
Conclusions Lung cancer nodule volumes can be accurately quantified using CT image data. Estimates deteriorate with increased slice spacing and thickness due to partial volume effects. More research is needed to find techniques that work well across thicknesses.