Information Technology Reference
In-Depth Information
with AIS. The inclusion criteria for this study was adolescent subjects who had their
X-rays taken during a scoliosis clinic consultation for either diagnosis or follow-up,
and had a calibration object placed during radiographic acquisition in order to
compare results. For each patient in the data set, a 3D reconstruction of the spine
was obtained from both methods. A series of clinical 2D and 3D geometrical
parameters were subsequently computed from these models and compared between
both techniques.
5.1.2 In Vivo Clinical Validation
Table
1
presents the results from this validation. Retro-projection errors are signif-
icantly lower (p
05) when using the proposed system based on self-calibration
with uncalibrated X-rays. It is somewhat dif
0
:
finding because
there is no gold standard to compare these results with. However because of the
intrinsic effect of 2D errors on the 3D model, it can be deduced that the better the
geometrical epipolar matching is in the stereoscopic vision, the better is the resulting
3D model. The value of the computerized Cobb angle in the frontal plane (C
PT
cult to evaluate this
PA
, C
MT
PA
,
C
PA
) with both systems is very similar, with slightly higher differences in the sagittal
plane (C
T4
T12
LAT
). The orientation of the planes of maximum curvature (
h
P
PMC
,
, C
L1
L5
LAT
h
MT
PMC
,
h
PMC
) offers very acceptable differences, with insignificant differences set at
p
05. Balance (y
T1
L5
, x
T1
L5
), however, gives a greater difference compared to
the previous linear appraoch, which is explained by the fact that the reference planes
are different. While the calibrated X-rays use the external calibration plate, the
uncalibrated X-rays use the images coordinate system as a reference.
0
:
Table 1 RMS difference and Wilcoxon test results of the geometrical indices measured on 60 3D
reconstructions of the spine obtained from the proposed system and a previous method [
28
]
Parameter
Symbol
Unit
Mean diff.
p-value
Epipolar error
e
mm
0.97
±
0.61
<0.001 (SD)
Cobb angle (PT)
C
PT
PA
deg
0.31
±
0.26
0.69 (NS)
C
MT
PA
0.19
±
0.17
Cobb angle (MT)
deg
0.59 (NS)
C
PA
Cobb angle (L)
deg
0.29
±
0.27
0.34 (NS)
Kyphosis
C
T4
T12
LAT
deg
0.52
±
0.41
0.16 (NS)
Lordosis
C
L1
L5
LAT
deg
0.63
±
0.37
0.21 (NS)
Max. deformity (PT)
h
P
PMC
deg
0.57
±
0.54
0.17 (NS)
Max. deformity (MT)
h
MT
PMC
deg
0.56
±
0.49
0.16 (NS)
Max. deformity (L)
h
PMC
deg
0.54
±
0.36
0.55 (NS)
h
MT
APEX
Axial rotation
deg
0.87
±
0.78
0.15 (NS)
Frontal balance
y
T1
L5
deg
0.97
±
0.53
0.04 (SD)
Sagittal balance
x
T1
L5
deg
1.88
±
1.15
<0.01 (SD)
