Repository-Residuum model shape and volume assessment

Data collection method:

In this study, ten residual limb models were scanned by three independent observers, each on three separate occasions, using two different scanners (i.e. 180 scans), over a 4 months’ period (May – August 2016). The models were selected from anonymous transtibial (n=5) and transfemoral (n=5) amputees to evaluate a large range of representative shapes and volumes. The models were distributed via prosthetics centres. They were manufactured using a standard carver for milling out foam models and a standard negative plaster-bandage wrap cast as a mould in which liquid plaster could set for plaster models.

Data processing and preparation activities:

Three independent observers were trained to use both the Artec Eva and Romer scanners. Prior to data collection, they completed two 2-hour familiarisation sessions, using both scanners. Ten different sessions were organised to measure each of the ten selected residual limb models. During each session, each model was measured three times by each observer with the two different scanners. This resulted in a total of 18 measurements (2 scanners × 3 observers × 3 repetitions) per model/session. The observers performed the measurements in randomised order (both observer sequence and scan sequence), with a 10-min break between each scan. Time per scan was between 1 and 3 minutes for the Artec Eva and Romer scanners, respectively. Prior to scanning, three 4 mm diameter hemispherical adhesive markers made of soft rubber were placed on the 3D surface of each residual limb model to approximately identify three anatomical landmarks for transfemoral (greater trochanter, Scarpa’s triangle and ischial tuberosity) and transtibial (tibial crest, fibula head and popliteal fossa) models. The distal borders of these markers were used to determine a plane used as the proximal end of each scan. Each model was placed on a metrology table for the Romer scans and on a normal table for the Artec Eva scans, with the distal end of the residual limb pointing upward. Artec Eva and Romer data files were processed using the same software used for data collection: Artec Studio 9.2 (Artec Group, Luxembourg, Luxembourg) and Geomagic Studio 2014 (Geomagic - 3D Systems, USA). Both the Artec Eva and Romer mesh models were exported and aligned manually in the same reference system x, y and z, using a graphical user interface according to the positions of the anatomical markers on the model. The volume of each residual limb model was calculated using the distal end to the proximal end of the residual limb as indicated by the plane and defined by the three anatomical markers. Parallel to this first plane 19 other planes were defined across the residual limb volume, obtaining a set of 20 parallel different sections at intervals of 5% across the residual limb length, with the first section (i.e. 0%) indicating the first proximal section of the residual limb model. For each section created by the 20 planes the relative Cross Sectional Area (CSA) and the perimeter (PE) were calculated. To assess residual limb model geometrical differences between the two scanners, the Root Mean Square Error (RMSE) between each pair of aligned scans was calculated. In addition, assuming the residual limb to be confined in a bounding box, residual limb sizes (width, depth and length) along three axes (x, y, and z, respectively) were calculated. Body Centre of Mass (BCOM) coordinates were calculated assuming the material of the models to be homogenous. For these calculations, each volume was processed using the Compute geometric measures filter in Meshlab software.

Technical details and requirements:

The use of high precision and resolution laser scanner has been suggested for evaluation of new scanning systems. For this reason, we used the Romer high precision and resolution scanner (Romer scanner, CMS108, Hexagon, UK) as the criterion measure to validate the ‘trueness’ of the Artec Eva scanner (practical measure). The Romer scanner is a powerful tool integrated with a Romer coordinate measuring arm that comprises different rotation axes to allow freedom of movement. It uses a laser line to reconstruct the 3D model with an accuracy of about 0.04 mm. In contrast, the Artec Eva (practical measure) is relatively small and uses regular flash bulb technology, illuminating the object with patterns of stripes by normal visible light to reconstruct 3D data from the surface with a reported accuracy of 0.5 mm.

Additional information:

STATISTICAL ANALYSIS: Accuracy of the Artec Eva scanner was assessed in terms of validity (trueness) and reliability (precision) using the statistical approach suggested by Hopkins, which was used to assess the validity and reliability of DEXA imaging methods. The following scanning variables were considered: residual limb volume, residual limb sizes (width, depth and length) along the three different axes (x, y, and z, respectively), BCOM coordinates, CSA and PE for each of the 20 levels of the residual limb length, where level 0 was defined as the plane passing through the distal border of the three anatomical/reference points. To ensure normality of the sampling distribution, each measurement was log transformed before analysis and back transformed after analysis. Log transformation was necessary to ensure uniformity of error, particularly where larger values of the original variable have greater absolute but a similar relative (%) error. Log transformation was not applied to the BCOM variables, since they were expressed in the relative Romer reference system.