Comparison of Image Generation and Processing Techniques for 3D Reconstruction of the Human Skull

Abstract

Although in the last few years, the use of the non-invasive medical techniques for diagnosis and treatment has experienced a huge development, mainly due to advancement in technology, for research and education these methods are still elaborate, expensive and not readily accessible. The purpose of our study was to compare the accuracy of an unconventional, non-invasive and relatively inexpensive Microscribe (3D digitizer) with a standard widely used and expensive CT-Scan and/or MRI for 3D reconstruction of a human skull, which will be used for biomechanics studies. Two models of the human skull were developed (reconstructed), one using the 3D coordinates generated by the Microscribe 3D digitizing unit and another one using the CT-Scans (2D cross-sections) obtained from a GE scanner. Using the hand-held digitizer, the Microscribe, X, Y and Z coordinates of a human skull were generated to create the first computer model. The 3D coordinates were brought as splines in to 3D Studio Max, a 3D modeling software, and U-lofted to form a solid NURBS model. The Microscribe captures the physical properties of a three-dimensional object and translates them into a 3D model. This kind of device is used to collect data directly from the surface of the study object. The stylus tip is moved over the contour of the object following its surface until the entire surface is digitized. Usually, points are drawn on the object’s surface in order to facilitate the digitizing process. 3D Studio Max takes this “raw” data and produces complex 3D models using various modeling techniques. For making the first skull model a technique called DRAW SPLINES was used. This method allows the user to begin a new spline or to do multiple splines by adding splines to those already created. I used this command to digitize my model because it is easy to use, quick and it gives the most accurate result. The final model was obtained in three steps: half of the skull was digitized and the first object was obtained, the MicroscribeSpline object (Fig. 1). The splines were transformed in NURBS curves and the second object was called NURBS Curves object. Finally, in the third phase, the NURBS curves were transformed in NURBS surfaces using the NURBS surface command, U-LOFT, and the final model, NURBS surface object, was obtained (Fig. 2). The entire skull was obtained from 2 identical halves of the same skull. The model was created using symmetry method because of the model’s organic complexity. The solid model was then exported to FEA software for analysis. (Fig. 3) The second skull model was created using the 2-D cross-sections obtained from the GE Helical Hi Speed - FX/i scanner (Fig. 4). The same skull used in the first part of the study, for modeling the first virtual model, was scanned following both sagittal and frontal planes. The interslice distance was set as being 3 mm. 48 CT slices for every analyzed plane were obtained. The CT cross-sections were captured as DICOM files using the E-film software and exported as TIFF images. The TIFF images were brought into OPTIMAS (image analysis software), which extracted the X, Y coordinates of each cross section using the POINT MORPHOMETRY option. A visual basic program was developed to convert the extracted coordinates to closed curves under Unigraphics SolidEdge software. To obtain the final model, the external boundaries of each cross section were lofted using LOFT PROTRUSION command. To find the best result, a second approach was developed in parallel using Adobe STREAMLINE and image processing software, which extracts the boundaries of each cross section and exports them as DXF files, compatible with the Solid Edge program. Both models were then subjected to stress analysis using Finite Element Analysis software. The analysis results obtained from the two scanning techniques will be discussed and presented, including the pros and cons of using the more accurate and expensive CT-scans versus the inexpensive hand-held scanner and their effects on finite element models. For this study, different image processing software such as OSIRIS, SCION IMAGE, EFILM, 3D DOCTOR, OPTIMAS and STREAMLINE were investigated in order to find the best interface to capture, reconstruct and model body data. The features, availability, cost and user-friendliness of these software tools will also be presented.