Processing Micro CT Bone Density Images

1
Processing Micro CT Bone Density Images

• NASA SHARP Student Paul Nelson NASA Mentor
  John DaPonte Ph. D.
• Team Members Michael Clark, Elizabeth Wood,
  Thomas Sadowski, Paul Thomas
• Southern Connecticut State University (SCSU),
  Computer Science Department
• Suny Stonybook University NASA SHARP Program

2
Project Overview
Original Image
After a few programs including deconvolution

• 6 sets of high resolution (10.5µm), 8 sets of low
  resolution (17.5µm) image stacks of approximately
  86 micro CT mouse images of trabeculae before and
  after exposure to weightlessness
• Iterative deconvolution, no pre-processing and
  Gaussian blur.
• Iterative deconvolution was qualitatively better
  then Gaussian blur and no pre-processing
• Provide quantitative verification of the
  qualitative findings through numerical evaluation
  of data obtained from a variety of programs

3
Current Work

• Areas of focus for this bone density study
  include the fractal dimension, entropy
  enhancement (EME), bone area, bone thickness, SMI
  , and BV/TV
• A majority of these parameters have already been
  collected for high/low resolution images

4
Calculations Background

• Bone Volume/Tissue Volume (BV/TV) Total Bone
  Volume
• 
  Total Tissue Volume
• Structure Model Index (SMI) 6x(SxV/S2)
• SChange in Surface Area Caused by Dilation,
  
  SObject Surface Area Before Dilation,
  VInitial Volume
• Bone Area Pixel Ratio (BAPR) of Bone Pixels In
  Threshold Image
• Total of Pixels
• Trabecular Thickness Pixel Ratio (TTPR) of
  Bone Pixels In Threshold Image
• of Bone
  Pixels In Skelonized Image

5
High/Low Res. Comparison Data
6
High/Low Res. Conclusions

• The average bone thickness for the low resolution
  images showed a slight thinning for deconvolution
  and greater thickening by the Gaussian
• The deconvolved data was closer to the data with
  no pre-processing
• For the high resolution images this feature is
  further shown because the Gaussian was thickened
  greatly and the deconvolved averages to about the
  same as the no pre-processing
• The average bone area for the low resolution
  image showed trends similar to average bone
  thickness
• This difference is not as large for the Gaussian
  as in the large for deconvolution in the high
  resolution images.

7
BV/TV Graphs and Conclusions

• For BV/TV the bone loss is lower for those mice
  that experienced more bone loss and higher for
  those that experienced less bone loss
• After suspension occurs, BV/TV always decreases
• The SMI appears inversely related to the BV/TV
  according to these graphs

8
Future Work

• To date, the fractal dimension values have yet to
  be compiled. Theses parameters will continue to
  be analyzed to identify any trends that might
  allow for the separation of more from less bone
  loss and before and after experiencing
  weightlessness.
• A planned future parameter of study is
  connectivity analysis.

9
The End