Skip to main content
. Author manuscript; available in PMC: 2019 May 1.
Published in final edited form as: IEEE Trans Med Imaging. 2018 May;37(5):1103–1113. doi: 10.1109/TMI.2017.2781541

TABLE IV.

Cartilage surface positioning errors (in mm) achieved by hierarchical classifier, gradient cost and single-stage RF classifier using 3D LOGISMOS algorithm. Paired t-test comparisons are reported for (NAF+RF vs. Gradient) and (NAF+RF vs. RF only)

n=108 NAF+RF Gradient p-value RF only p-value
Femur signed −0.01±0.18 −0.31±0.24 ≪ 0.001 −0.10±0.17 ≪ 0.001
Femur unsigned 0.55±0.11 0.69±0.13 ≪ 0.001 0.56±0.10 ≪ 0.001

cMF signed −0.04±0.29 −0.38±0.58 ≪ 0.001 −0.11±0.27 ≪ 0.001
cMF unsigned 0.52±0.16 0.78±0.35 ≪ 0.001 0.55±0.17 ≪ 0.001

cLF signed −0.26±0.24 −0.52±0.35 ≪ 0.001 −0.36±0.20 ≪ 0.001
cLF unsigned 0.42±0.12 0.65±0.20 ≪ 0.001 0.47±0.11 ≪ 0.001

Tibia signed 0.06±0.17 −0.11±0.35 ≪ 0.001 0.11±0.22 ≪ 0.001
Tibia unsigned 0.60±0.14 0.79±0.20 ≪ 0.001 0.62±0.18 ≪ 0.001

cMT signed −0.15±0.31 −0.25±0.77 0.193 −0.09±0.34 0.003
cMT unsigned 0.52±0.20 0.92±0.41 ≪ 0.001 0.58±0.22 ≪ 0.001

cLT signed −0.03±0.32 0.36±1.12 ≪ 0.001 −0.01±0.32 0.16
cLT unsigned 0.46±0.17 0.79±0.99 ≪ 0.001 0.47±0.18 0.31
*

Bold indicates statistically significant reduction in positioning errors.