How to measure children’s feet: 3D foot scanning compared with established 2D manual or digital methods

Participants

The study was approved by the institution’s Ethical Committee of the University of Potsdam (No. 04/2018). Children were recruited from local primary schools as well as local children shoe stores. Children’s parents were informed about the purpose and applied methods and gave informed consent before children’s voluntarily participation in the study. Based on inclusion (age 5-10 yrs) and exclusion criteria (no acute/chronic pain and/or injury at the locomotor system), measurements were performed on n = 297 children. For final analysis n = 277 children (125 m / 152 f; mean ± SD: 8.0 ± 1.5yrs; 130.2 ± 10.7cm; 28.0 ± 7.3kg) could be considered since 20 presented incomplete data sets (17 missing values for 3D scans; 3 missing values for 2D scans) and were therefore excluded. Table 1 shows detailed information (including mean and 95% confidence interval) on anthropometric data of the included participants.

Table 1 Anthropometric data of included participants (mean; 95% Confidence interval (CI); range)Experimental protocol

After parents’ informed consent to the study was given, information on age and sex were collected as well as anthropometric measurements (body height (cm) measured with seca mobile stadiometer 217; body weight (kg) measured with seca flat scale 899, seca Germany) were conducted barefoot while participants remained wearing all-day clothes. Afterwards, geometry of the right foot was measured in static condition (stance) with three different measurement systems (fixed order): (1) manual foot measurement [1], (2) 2D foot scanning on a 2D desk scanner and (3) 3D foot scanning by use of a 3D hand-held scanner [6]. All foot measurements were performed in a bipedal stance. The same experienced examiner and two scientific assistants performed data acquisition throughout the whole study. The whole measurement procedure took about ten minutes per child.

Manual Foot Measurements (MF)

Static foot length (FL; [mm]) and (projected) fore foot width (FW; [mm]) were measured in a standing position with a standardized foot measuring device developed to determine shoe size (WMS® foot measurement system with an attached millimeter scale, DSI, Offenbach, Germany; Fig. 1A) [1, 16]. The device consists of a base plate, a rear margin, and a separator (left/right foot) orthogonal to the rear margin. The feet were set left/right to the separator having contact on the medial foot side and with the heel all the way back to the rear margin. Foot length and width are measured with a front and side slider (orthogonal to separator/rear). Therefore, foot length is defined as maximum heel to longest toe distance and foot width is the maximum forefoot width. In addition, the anatomic foot ball breadth (Fig. 1B) was measured with a tapeline [mm] around the big toe joint and the small toe joint in a standing position with half-body weight bearing on each foot (Fig. 1B). All outcomes are detailed in Table 2.

Fig. 1figure 1

Devices and setups for the Manual Foot Measurements (A/B), 2D Foot Scan (C) and 3D Foot Scan (D/E). A standardized foot measuring device (WMS® foot measurement system, DSI, Offenbach, Germany). B setup for manual foot ball breadth measurement. C 2D desk scanner (iScan 2D, IETEC Biomechanical Solutions, Germany)). D 3D foot scan measurement setup. E 3D hand-held light scanner (Artec Eva; Artec Group, Luxembourg)

Table 2 Outcome measures for all three measurement methods2D Foot Scan (2D)

To scan the children’s right foot, the participants were instructed to stand still on a 2D desk scanner (iScan 2D, IETEC Biomechanical Solutions, Germany; Fig. 1C) in a hip wide bipedal stance with weight equally distributed across both feet. Shoes and socks were removed before the measurements. During the scan procedure, a lightproof blanket was placed over the feet and the scan bed to ensure optimal light conditions for the scanning procedure. Afterwards, the examiner visually inspects the scanning result. In case of reduced scan quality (e.g. foot placed too close to the boarders of the scan platform; blurry image due to movement artefacts), the scanning procedure was repeated. The foot scans were recorded within A3 size and a resolution of 96 dpi (= 1 Dot per Inch = 25.4 mm).

Image processing: Each foot scan was analyzed using a custom-made image recognition software (FeetAnalyzer, MATLAB 2018). This software allowed a semi-automatic detection of the landmarks for the calculation of the main outcome measures. Every automatic detection was controlled through visual inspection by an experienced examiner. If automatic detection failed (e.g. due to movement artefact), the investigator applied manual correction (<2% of all cases analyzed). Main outcome measures for 2D scans are foot length (FL; [mm]) and five parameters of foot width: projected (fore)foot width (FW_P, [mm]), anatomic (fore)foot width (FW_A), (technical) foot ball width at 65% foot length (FW_65, [mm]), (technical) foot instep width at 50% foot length (FW_50; [mm]) and (technical) heel width (HW; [mm]) at 20% foot length. All outcomes are detailed in Table 2 and visualized in Fig. 2.

Fig. 2figure 2

Visualization of all outcomes of the 2D and 3D Foot Scan. A 2D Foot Scan Outcomes. B 3D Foot Scan Outcomes

3D Foot Scan (3D)

Three-dimensional foot scans were executed on a platform (height: 120 cm) bordered by a handrail and accessible by a stair. Children stood still in a slightly lunge position with the right foot in the front and at the edge of the platform (Fig. 1D). Children were instructed to stand still and look straight forward (not downward) with hands fixing the handrail. During measurement, an experienced examiner scanned the right foot by driving the hand-held 3D light scanner (Artec EVA, Artec Group, Luxembourg) [6, 17] around the foot beginning at the lateral heel side over the toes leading to the medial heal side. In case of foot movement artefacts, the measurement was repeated. Scanning was performed at a speed of 16 frames per seconds and the depth of the scanning field was adjusted to 400 mm and 100 mm. The 3D accuracy of the scanner is up to 0.05 mm, and the 3D resolution is up to 0.1 mm. During scanning, a laptop with Artec EVA Studio software (version 9.2.3.15; Artec Group) was used to host the 3D scanner.

Image processing: The 3D scan files were processed with an industry-acclaimed software package for advanced 3D scanning and data processing (ARTEC Studio V11; Artec Group, Luxembourg). For each scan file a 3D model of the foot was created including the following steps: visual inspection of the scans to check for possible scan errors (e.g., cramped or lifted toes, defective objects in the scan area, noise etc.), scan alignment, global data registration, fusion of data into a 3D model (Fusion), and final editing of the 3D model (e.g. cut to size). Thereafter, each scan was oriented manually according to the established coordinate system (y-axis = longitudinal foot axis from heel to second toe). The origin of the coordinate system was set to the maximum arch point of the heel. Afterwards the model was exported to STL format. Subsequently, the STL file was imported into a custom-made analysis software for 3D foot model (FootOpenGL, PFI, Germany). To obtain the information for the desired main outcome measures, defined measurement points, cuttings and lines are added to the model (Fig. 2). The used software consisted of an automatic detection mode for the points and cuttings for the defined main outcomes measures. Every automatic detection was controlled through visual inspection by an experienced examiner. If automatic detection failed, the investigator applied manual correction (<2% of all cases analysed). Main outcome measures for 3D scans are foot length (FL; mm]), foot ball breadth (FB; [mm]) and four parameters of foot width: projected (fore)foot width (FW_P; [mm]), anatomic (fore)foot width (FW_A), (technical) foot instep width at 50% foot length (FW_50; [mm]) and (technical) heel width (HW; [mm]) at 20% foot length. All outcomes are listed in Table 2 and visualized in Fig. 2.

Data analysis and statistics

Initially, data was recorded in a case report form (CRF) followed by transforming and saving in a statistical database (JMP Statistical Software Package 14, SAS Institute®). All measures were checked for plausibility (e.g. range check, age: 5-10 years; body height: < 2.00 m; body weight: < 120 kg). Implausible values were recalculated or corrected as documented in the handwritten CRF. Otherwise values were erased out of the database. Statistical analysis was performed descriptively calculating mean and standard deviation followed by inferential statistics. All outcomes were checked for normal distribution with Shapiro-Wilk-Test. Following this, analysis of variance (e.g. one-way repeated-measures ANOVA; paired t-test) for dependent samples was applied to test for differences between measurement methods. In case of significance, Tukey-Kramer-Test was applied for post-hoc analysis. Level of significance was set to α = 0.05. To account for multiple testing, the level of significance was adjusted (Bonferroni correction) to α = 0.008. In addition, Bland and Altman analysis including bias and 95%-limits of agreement (LoA) were calculated to evaluate reproducibility between the methods used (manual vs. 3D; manual vs. 2D; 2D vs. 3D) [18, 19]. Further, heteroscedasticity was analyzed calculating Pearson correlation (for the mean value of two measurement approaches and the difference between the two measurement approaches).

留言 (0)

沒有登入
gif