Accuracy evaluation of 3D reconstruction of transfemoral residual limb model using basic spline interpolation

Abstract

In this paper, a study on the accurateness of 3D reconstruction for transfemoral residual limb based on Magnetic Resonance Imaging (MRI) using basic spline (B-Spline) interpolation feature is presented. Many researches have constructed 3D models by using the Non-Uniform Rational B-Spline (NURBS) approach; however, almost none of those studies have elaborated in detail the methodology used in the project and the accuracy of the model’s volume against the residual limb volume. This study focuses on the optimization of the residual limb model’s volume by investigating the effect of spline points on the volume value. This project is divided into 3 phases, namely pre-segmentation, segmentation, and 3D reconstruction. However, this study focuses on the segmentation phase in which four different spline point numbers are chosen, namely 72, 36, 24, and 12 points; CAD software CREO Parametric (PTC) was used for this process. A higher number of spline points achieved greater accuracy for the model’s volume. The volume (mm3) of the 3D model with the specified number of spline points was evaluated by comparing it with the volume of the 3D model created by using image processing tools from MATLAB (Math Works). The results show the increment in accuracy of the volume value when the number of spline points is increased. The highest accuracy in terms of volume value is achieved when the model is created with 72 and 36 spline points, with an average error percentage of 6.36%, which realized the hypothesis. The results indicate that CAD software, which is a technical drawing tool, can also be used in the biomedical field to design 3D models of the human anatomy with high accuracy, if the software includes a spline interpolation feature.