Case Study - Skull stripping and image segmentation

These algorithms are described in detail in the Software Section and will not be discussed here. These analyses were used to study reliability of the algorithms for repeated measures of the same individual within 2 weeks and comparison between 10 AD patients and 10 cognitively normal controls.

Measure

TCV

Brain

Gray

White

CSF

Time 1 cc

1371.4 ± 150.2

1069.6 ± 106.5

675.3 ± 74.8

394.3 ± 44.4

301.9 ± 54.5

Time 2 cc

1355.8 ± 141.0

1055.5 ± 102.0

665.3 ± 77.2

390.2 ± 59.7

289.2 ± 51.5

Time 1 %TCV

---

78.1 ± 2.1

49.3 ± 1.9

28.8 ± 2.2

21.9 ± 2.1

Time 2 %TCV

---

77.9 ± 2.5

49.1 ± 3.1

28.8 ± 3.6

21.3 ± 2.2

ICC

0.99

0.97

0.95

0.92

0.90

Results of the repeated analyses are displayed in the table above. Data are presented in both cubic centimeters as well as the proportion of intracranial volume to correct for the effect of gender. While there were subtle mean differences in TCV estimated from the skull-stripping algorithm (1.1%), all repeated measures were found to be highly reliable, and when corrected for the differences introduced by the skull strip algorithm, were reduced to 0.33% on average with a range of 0-0.6%.

Group differences in brain and tissue type volumes are shown in the figure above. Significant group differences between AD and cognitively normal individuals were found for total brain, gray matter and CSF volumes.

 

Right Hippocampus

Left Hippocampus

Total Hippocampus

Normal

0.138 ± 0.020

0.132 ± 0.018

0.270 ± 0.036

AD

0.109 ± 0.024

0.099 ± 0.028

0.208 ± 0.051

p-value of difference

0.0083

0.006

0.0053

Results of hippocampal measures are seen in the table above. Significant findings are again seen with these measures. The findings of significant group differences using a sample of 10 mild AD patients supports the accuracy of these methods to detect biological differences.

Up (Case Studies)