This methodological study explored patients included between January 2020 and September 2020 in a local cohort and was approved by local ethics committee of the University Hospital of Bordeaux (CEREBDX 2021 – 31). Every patient gave a written consent in accordance with the Declaration of Helsinki. This study followed the GRRAS guidelines.

Patients monitored for camptocormia in the orthopedic medical unit of the University Hospital of Bordeaux by a physician specialized in spinal disorders, were proposed for inclusion.

Inclusion criteria were: (i) aged more than 18 years-old and (ii) presenting with camptocormia according to usual clinical criteria8 (some degrees of forward flexion of trunk during walking that disappeared in the recumbent position), and (iii) having a recent (<15days) EOS examination performed in ambulatory care.

Exclusion criteria were: (i) inability to stand or walk three minutes without the help of another person or mobility equipment and (ii) refusal to participate in the study or inability to sign written consent.

SAMD3D is a non-invasive and non-irradiating measuring tool. It has been developed by the company SAM Research, which allows performing postural assessments in 3 dimensions. The SAM3D device is a mechano-optical assembly coupled with a specific software producing image acquisitions in 3D. The function of 3D reconstruction is a standard model using the fundamental triangulation principle of any 3D reconstruction from images described by Fig. 1.

Fig. 1.

SAMD 3D device which allows to measure posture based on specific landmarks. The principle of 3D reconstruction is a standard model using the fundamental triangulation principle of any 3D reconstruction.

(0.44MB).

The acquisition system includes two high-definition image sensors (4K CMOS, 14MP, 40ms exposure time, 1mm accuracy), coupled with the “back-side-illumination” technology, a roll-up (120cm x 200cm), and a floor plate (120cm x 120cm). The two cameras are fixed one above the other 10cm apart, allowing optimal precision/distance performance. The synchronization of the two images captured by the cameras is directly ensured by the software.

Automatic detection was performed on anatomical landmarks indicated by the operator. Landmarks include a total of 18 points distributed on the back, which reflects the inflexion points of the spine, including the wrists and heels (Fig. 1).

A complete stereo camera calibration is realized during the installation of the system SAM3D.31,32 A daily-basis check was done on background task to verify if the camera intrinsic and extrinsic had changed. With the visual landmarks of which the ground-truth are known, the system can auto-calibrate its extrinsic parameters. A well calibrated SAM3D camera has a reprojection error of 0.2 pixel and the 3D reconstruction RMSE (root-mean-square error) are estimated as 0.35 mm on the X axis and 0.61 mm on the Y axis, and the depth error on the Z axis is 0.87 mm.

First, the patients were asked to undress to their underwear and remain barefoot. Then, the examiner used a dermal pen to label the 18 predefined anatomical landmark points that would enable the automatic detection to establish a 3D topography after image processing. The subject was positioned about 3.5m from the image sensor (Fig. 1). Because a scoliosis is often associated with camptocormia, the measurements were performed in 3 dimensions: (i) the frontal plane, (ii) the sagittal plane, and (ii) the transverse plane. SAM3D allows reconstruction of axes and selected anatomical segments.

In our study, we decided to explore the following sagittal parameters:

• -

  The cervical arrow in C7 (mm) to quantify trunk forward flexion in unit of length. This is the distance between the 2 vertical lines, the first passing through the spinous process of C7 and the second through the middle of the postero-superior iliac spines (PSISs) segment (Fig. 2).

  
  

  Fig. 2.

  Sagittal view of the landmarks analysis. A) bone schematic view, B) view of a healthy control (compared to the Gaussian distribution: in yellow > 1SD; in red >2SD), C) view of a patient with camptocormia, D) arrow in C7 (patient with camptocormia).

  (0.59MB).
• -

  The angles of kyphosis and lordosis: Kyphosis and lordosis were calculated based on the tangent of the inflection points of the spine curves in the cervico-thoracic and thoraco-lumbar areas (Table 1).

  Table 1.

  Average measurement of the arrow at C7 made by SAM 3D.

  	arrow at C7 (mm)	Lin's CCC	Kyphosis angle (degrees)	Lin's CCC	Lordosis angle (degrees)	Lin's CCC	Knee flexion (degrees)	Lin's CCC
  SSP (mean ± SD)
  User 1	117 ± 59		48 ± 15		27 ± 17		6 ± 4
  		0.97 (95 %CI 0.95, 0.99)		0.94 (95 %CI 0.87, 0.97)		0.89 (95 %CI 0.80, 0.95)		0.88(95 %CI 0.76, 0.94)
  User 2	119 ± 60		50 ± 16		29 ± 20		6 ± 4
  USP (mean ± SD)
  User 1	80 ± 60		42 ± 13		33 ± 18		5 ± 3
  		0.98 (95 %CI 0.96, 0.99)		0.94 (95 %CI 0.88, 0.97)		0.90 (95 %CI 0.80, 0.95)		0.88 (95 %CI 0.77, 0.94)
  User 2	78 ± 57		42 ± 14		32 ± 17		5 ± 3
  WSP (mean ± SD)
  User 1	117 ± 70		51 ± 17		30 ± 15		6 ± 3
  		0.97 (95 %CI 0.95, 0.99)		0.89 (95 %CI 0.78, 0.94)		0.92 (95 %CI 0.83, 0.96)		0.89 (95 %CI 0.79, 0.95)
  User 2	114 ± 73		52 ± 16		31 ± 15		6 ± 3

  PSISs, postero-superior iliac spines; SD, standard deviation; SSP, spontaneous standing position; USP, upright standing position; WSP, walking- like position;
• -

  The leg inclination: the angle between the vertical and the segment “posterior edge of the calcaneum – posterior edge of the popliteal fossae” which shows knee flexion (Fig. 2).

The measurements were performed in the 3 best positions reflecting the dynamic phenomenon of camptocormia: (i) Spontaneous Standing Position (SSP) (defined by the spontaneous standing of the patient without instruction), (ii) Upright Standing Position (USP) (defined by the standing position when the patient was asked to be as straightened as possible), and (iii) Walking-like Standing Position (WSP) (asking the patient not to stand up straight after a short walk inducing trunk inclination).

Two separate operators, trained physicians in rehabilitation medicine, blinded to each other, applied and removed dermal landmarks then acquired the images in all 3 positions after explaining and showing each position to the patient. One out of the rater was familiarized with the tool when the other was naïve and never used SAM3D before the study. The completed time for the acquisition of the 3 positions’ images per patient took less than 5 minutes, including for the application and removal of the landmarks. From each acquisition, a 3D model from the selected axes and anatomical segments was reconstructed after thorough validation of the automatic anatomical landmark locations by the operator. Then the different parameters were computed automatically.

The Sagittal vertical axis (SVA) using EOS system was collected to put into perspective SAM3D results with those from a radiographic gold standard. These data resulted from an out-patient clinical routine examination (less than 15 days), thus the EOS procedure was not controlled.

Sample calculation: at least 31 patients for Lin's CCC > 0.81 and expected at 0.92 with a power of 0.80. The interrater reliability of measurements has been assessed by the Lin's Concordance Correlation Coefficient (CCC).33

According to Partik and al,34 a CCC between 0.51–0.60 is considered to be poor, between 0.61–0.70 slight, between 0.71–0.80 moderate, between 0.81 - 0.91 fairly good, between 0.91–0.95 very good, and >0.95 excellent. To explore changes between the 3 positions, we analyzed the measurements of the arrow in C7, and performed ANOVA with repeated measures and contrasts. We also explored the reliability (Lin's CCC) between arrow in C7 obtained by the SAM3D system in the 3 positions for each examiner and C7-SVA using EOS system. A p-value < 0.05 was considered significant.