Short communication
3D reconstruction of the ribs from lateral and frontal X-rays in comparison to 3D CT-scan reconstruction

https://doi.org/10.1016/j.jbiomech.2007.09.034Get rights and content

Abstract

Subject-specific three-dimensional (3D) reconstructions of the ribs can be obtained from biplanar X-rays. The goal of this study was to evaluate the accuracy and the inter-observer reproducibility of this technique in comparison to CT-scan reconstructions. CT scans and biplanar X-rays were obtained from 50 ribs (from three cadaveric rib cages). Three experienced experimenters reconstructed each rib from biplanar X-rays. Morphometric parameters were then computed from the rib midlines. Differences were computed between parameters obtained from the 3D reconstructions based on biplanar X-rays and from CT scans. The accuracy was computed as the mean of this difference for the 50 ribs from all three experimenters. The inter-observer variability was assessed using the coefficient of variation (CV) between the three observers. The CT-scan reconstructions were considered to be the gold standard in spite of their limitations for rib reconstructions. According to the different linear parameters, the accuracy of the reconstructions was found to be between −6 mm (−2%) and 3 mm, (4%). The accuracy of the current method was close to that of CT-scan reconstructions. The inter-observer variability was between 3% and 6%. Frontal and lateral X-rays are commonly obtained clinically, so 3D reconstructions can be used without increased radiation exposure to the patient.

Introduction

A three-dimensional (3D) patient-specific geometrical model of the ribs is essential for clinical evaluation in scoliosis (Dubousset et al., 2003; Labelle et al., 1995), a complex 3D deformation of the trunk. Patients with scoliosis often require conservative intervention in the form of orthotic braces. In these patients, information about rib shape and length asymmetry would provide a better understanding of mechanisms underlying the relationship between the rib curvature and vertebral rotation (Erkula et al., 2003), leading to a better brace design. Patient-specific biomechanical finite element models have been developed to predict the outcome of orthopaedic treatment. These models require ribs geometry (Aubin et al., 1995; Le Borgne et al., 1998). The 3D information could be obtained by CT scan, but with a high-radiation dose for the patient. In the specific case of the ribs, a method based on two frontal X-rays (conventional frontal and frontal, with a 20° angle from the horizontal) was previously developed (Dansereau and Stokes, 1988) and evaluated to assess in vivo geometric measurement variability (Labelle et al., 1995). A method based on three X-rays (same to as Dansereau and Stokes (1988) plus a lateral one) was used to model the spine, ribs (Cheriet et al., 2007) and the spine, ribs and pelvis in 3D (Delorme et al., 2003). In these previous studies, the incidence angles of the X-rays were not the ones most often used clinically (i.e., frontal and lateral). In order not to add unneeded radiation for the patient and facilitate translation of this work to the clinic, 3D reconstructions based on the clinically obtained X-ray incidences (frontal and lateral) were proposed.

3D reconstruction methods based on the two X-rays (frontal and lateral) were developed for the spine (Aubin et al., 1997; Mitton et al., 2000; Pearcy, 1985; Pomero et al., 2004), the knee (Laporte et al., 2003) and, more recently, the rib cage (Benameur et al., 2005; Laporte et al., 2004). The assessment of the rib reconstruction method proposed by Benameur et al. (2005) was limited because the results were compared with those by a previous stereoradiographic method (Dansereau and Stokes, 1988) because ribcage CT data were not available (Benameur et al., 2005).

The aims of this study were (i) to estimate the accuracy of the method presented by Laporte et al. (2004) by comparing 3D biplanar reconstructions with CT-scan reconstructions, and (ii) to evaluate the inter-observer reproducibility of this method.

Section snippets

Materials and methods

Three fresh-frozen human cadaveric thoraxes (age 75, 84, and 92 yrs) were obtained from the Mayo Foundation Anatomical Bequest Program (Rochester, MN, USA). The specimens were thawed at room temperature overnight. Fifty ribs from the three thoraxes (ribs, sternum and vertebrae) were used. Ribs 2–10 were used from two specimens and ribs 4–10 from the third specimen.

Results

Reference values obtained from CT scans were compared with parameters computed from the stereo-radiographic reconstructions and the differences for the three observers determined (Table 1).

The mean differences were for the length of the rib mid-line −6 mm [−2%], the maximum width 3 mm [4%], the chord length 0 mm [−0.2%] and the enclosed area 370 mm2 [3%] (difference=stereo-radiography–CT scan).

Regarding the inter-observer variability (computed from the three stereo-radiographic reconstructions),

Discussion

The relative mean differences were either positive or negative depending on the parameter considered. These differences were not only due to the stereo-radiographic technique but also done to the CT-scan reconstruction. The anterior limit of the rib is especially hard to define even in the CT scan. Therefore the CT scan, considered as a gold standard for the in vivo 3D reconstruction, is not an absolute reference. Thus, this study only leads to an approximation of the accuracy of the

Conflict of Interest

The authors have no conflict of interest for this study.

Acknowledgement

The authors wish to acknowledge M. Martinet and M. Chauvet for their important technical support.

References (22)

  • J. Dubousset et al.

    A new 2D and 3D imaging approach to musculoskeletal physiology and pathology with low-dose radiation and the standing position: the EOS system

    Bull. Acad. Natl. Med.

    (2005)
  • Cited by (33)

    • X-ray image calibration and its application to clinical orthopedics

      2014, Medical Engineering and Physics
      Citation Excerpt :

      Moreover, the usage of simple scaling objects cannot be used to overcome the loss of depth. Therefore, more sophisticated approaches have been proposed in literature combining X-ray image calibration with 2D–3D reconstruction techniques [4–7]. 3D reconstruction of the scoliotic spine was proposed by Novosad et al. [4].

    • Robust femur condyle disambiguation on biplanar X-rays

      2012, Medical Engineering and Physics
    • Reliability of a new method for lower-extremity measurements based on stereoradiographic three-dimensional reconstruction

      2012, Orthopaedics and Traumatology: Surgery and Research
      Citation Excerpt :

      Specially designed software included in the workstation allows three-dimensional (3D) modelling of the bone envelope and automatic calculation of specific clinical variables (Fig. 1). Biplanar stereoradiography and personalized modelling of the skeleton have been extensively developed for various anatomic regions including the lumbar spine [10], cervical spine [11], ribs [12], pelvis [13]. In a recent study, the use of this technique on lower extremities was validated on children [14].

    View all citing articles on Scopus
    View full text