Range data pre-processing for the evaluation of torso shape and symmetry in scoliosis

Abstract

Pre-processing range scans of the human torso for evaluating shape and symmetry changes in scoliosis are non-trivial. First, stray points from surrounding artefacts are often arbitrarily positioned and not amenable to automatic removal. Second, the asymmetrical alignment of the arms and neck makes cropping them difficult. Third, despite a plethora of methods, removal of holes by surface approximation for this niche application remains a challenge particularly in obscure regions like the sides and armpits. This paper proposes a novel surface approximation method and incorporates it into an integrated procedure for pre-processing range scans of the torso that includes interactive tools for cropping stray points and extremities. The new method, spline-fitted moving least squares (MLS), makes use of the Bezier curve and MLS algorithms. Numeric and clinical tests on scans of 30 volunteers, with and without scoliosis, show that the proposed method outperforms its constituent methods and a commercially available graphics package for this application.

Keywords:

• deformity
• moving least square
• scoliosis
• splines
• range scans

Notes

^1. FarField Technology Limited, Christchurch, New Zealand. FastRBF is a commercially available surface reconstruction package based on radial basis functions.

^2. INSPECK, Inc., Montreal, Quebec, Canada.

^3. KONICA-MINOLTA Photo Imaging, Inc., Mahwah, NJ, USA.

^4. This is achieved by working with the cross-sections in an increasing order of position. For all cross-sections apart from the first two, the previous and next two sections serve as a 3D ‘ribbon’ that provides a rough estimate of its shape.

^5. The distribution of distances between neighbouring points for this application was found to be skew-symmetric.

^6. Key control points contribute most to the shape of the curve represented by P.

^7. The degree of misalignment of the left and right extremities is an index of torso asymmetry.

^8. This assumption does not hold for arbitrary shapes. The mD can also be defined as twice the farthest distance of any point from the centre of the cross-section.