The pelvis is an essential element in the study of scoliosis since it
constitutes the base of the spine and its orientation may affects post
ural balance. In order to study the role of the pelvis in the evolutio
n and treatment of this disease, a new technique for the 3D personalis
ed reconstruction of the pelvis was developed. It consists in identify
ing and digitizing 19 pelvic anatomical landmarks on posters-anterior
and lateral x-rays and to reconstruct them in 3D with two techniques:
the DLT algorithm developed by Marzan (1976) and, for 6 of the 19 land
marks, an adaptation of it called DLT with confidence coefficients. Th
e latter takes into account the confidence given to the identification
of the landmarks on each x-rays. Two methods were used to validate th
e reconstruction of the pelvis, The first one, used for 11 scoliotic p
atients and ? dry pelvis specimens, consists in applying the reconstru
ction algorithm in an inverse way on the 3D coordinates of the reconst
ructed landmarks to obtain their 2D retroprojection on the x-ray plane
s, and thus comparing the retroprojected coordinates with the 2D digit
ized coordinates. The second method consists in measuring a dry pelvis
specimen and comparing the 3D measured landmarks with the ones recons
tructed with the x-rays of this specimen, For the first validation, re
sults have shown that the lowest retroprojection errors (less than 2.5
+/- 2.6 mm) for the scoliotic patient group are located on the superi
or base of the sacrum, on the sacral curve and on the acetabula, while
the highest (6.4 +/- 7.2 mm) were on the iliac crests. For the dry sp
ecimens, the retroprojection errors were below the millimeter. The sec
ond validation method showed 3D differences of 2.4 +/- 1.2 mm between
measured and reconstructed landmarks of a dry specimen, which is of th
e same order of magnitude as what is reported in the literature for ve
rtebrae. The reconstruction of the pelvis is thus considered adequate
and its graphical wireframe representation allows to visualise and mea
sure clinical indices concerning its orientation in space. Moreover, t
he reconstructed landmarks will be used to develop a personalised geom
etrical and mechanical model of the pelvis which, when integrated with
the one for the spine and rib cage, will allow to simulate in a more
realistic manner the biomechanical behaviour of the scoliotic trunk, p
articularly for the study of orthopaedic treatments with braces leanin
g on it.