Semi-automatic measurements of foot morphological parameters from 3D plantar foot scans

Background Foot healthcare research is focusing increasingly on personalized orthotic and prosthetic devices to address patient-specific morphology and ailments. Customization requires advanced 3D image processing tools to assess foot and leg geometrical parameters and alterations. The aim of this study is to present a new software for the measurement of the foot shape from 3D scans of the foot plantar surface. Methods A Kinect-based scanning device was used to acquire the 3D foot shape of 44 healthy subjects. A software was developed in Matlab to measure the foot main morphological parameters from foot scans. Principal Component Analysis was used to orientate the foot scans with respect to the same reference system. Accuracy, via percentage errors and Bland-Altman plots, and correlation of the software-based foot parameters were assessed against manual measurements. A normalized Arch Volume Index (nAVI) was proposed and correlated to the traditional Arch Index. Test-retest Intraclass Correlation Coefficient was used to assess the inter-session repeatability of foot measurements. Results The average percentage error between software and manual measurements was 1.2 +/- 0.8% for foot length, 9.1 +/- 3.7% for foot width, 22.3 +/- 13.5% for arch height and 23.1 +/- 12.7% for arch depth. Very strong correlations were observed for foot length (R = 0.97) and foot width (R = 0.83), and strong correlations for arch height (R = 0.62) and arch depth (R = 0.74). nAVI was negatively correlated to the Arch Index (R = -0.54). A small difference was found between software and manual measurements of foot length (Delta = 0.92 mm), a software overestimation of foot width (Delta = 8.6 mm) and underestimation of arch height (Delta = -1.4%) and arch depth (Delta = -11%). Moderate to excellent repeatability was observed for all measurements (0.67-0.99). Conclusions The present software appears capable to estimate the foot main morphological parameters without the need for skin markers or for identification of anatomical landmarks. Moreover, measurements are not affected by the foot orientation on the scanning device. The good accuracy and repeatability of measurements make the software a potentially useful operator-independent tool for the assessment of foot morphological alterations and for orthotics customization. nAVI may be used for a more realistic classification of foot types when 3D foot images are available.