The resin embedded skull (HUJ-PAL 3659) presents the ventral surface exposed on an epoxy resin plaque, retaining traces of the original limestone matrix on the surface. The epoxy resin holding the specimen is approximately 17.5 mm thick, 75 mm wide, and 195 mm long. The skull was scanned at the University of Texas at Austin High-Resolution CT Facility (tube voltage 150 kV, 0.16 mA, no filter, air wedge, 190% offset, slice thickness of 0.24 mm, S.O.D. of 130.0 mm, 1800 views, 1 sample per view, interslice spacing of' 0.2 mm, field of reconstruction is 70.0 mm, reconstruction offset 600, reconstruction scale 75). This resulted in 308 transverse slices at ~137 micron interpixel resolution and a slice thickness of 240 microns; each slice was saved as a 512 X 512 pixel tiff image in both 16 bit and 8 bit modes (Appendix 1). The sequential 308 slices together form a three-dimensional matrix, and each pixel representing a volume element (voxel) in a three-dimensional framework. Variations in the value of each voxel represent variations in relative X-ray attenuation, which closely mirror compositional variations. X signifies the transverse axis, Y the dorsoventral axis, and Z the longitudinal axis of the CT data set. Voxblast version 3.0 (VayTek 2000) was used to build rendered and lighted isosurface and sectional reconstructions. Voxblast was also used to resample the original 308 slice stack in the XZ plane, in order to isolate the 93 slices that contain the specimen–thus reducing the file size and memory requirements for processing, and providing better sectional illustrations for comparison with isosurface reconstructions. ImageJ version 1.32i (Rasband 2003) was used for analysis and tracing of slices and Adobe Photoshop 6.0 (Adobe Systems, Inc. 2001) for cropping and rotation of the original data set as well as increasing voxel density using the bicubic method of resampling. The latter process was used to optimize the isosurface reconstruction abilities of Voxblast, providing a more voxel-dense volume, thereby minimizing aliasing effects.
CT scanning and computer reconstruction have resolving limitations. These are manifested in spatial resolution, X-ray attenuation differentiation, noise artifacts, and reconstruction artifacts. Specimen geometry greatly influences both the available spatial resolution and artifacts caused by variation in X-ray scatter and differential absorption due to the amount and composition of material intersected by the X-rays at different angular rotations. Ideal specimens are cylindrical and small enough to sit close to the X-ray source, providing magnification via the fan beam projection onto the X-ray detectors. In the case of HUJ-PAL 3659, specimen size and geometry were less than optimal, presenting a long rectangular cross section, with the area of interest small relative to the block size, thereby reducing beneficial magnification. Serial sections do not clearly delineate individual elements but show centers of greater attenuation, allowing elements to be traced through the volume. Additionally, software problems encountered during the scanning process forced multiple interruptions and restarts, resulting in slice misalignment in the anterior and midsection of the skull and vertebral column. These misalignments are visible in the isosurface reconstructions as transversely oriented lines on the anterior portion of the skull and vertebral column. A loss of data in the midsection of the skull occurred during one of the scan interruptions, constituting approximately five slices and amounting to about a 1 mm gap, and is represented by the thick black transversely oriented line across the midsection of the skull in the CT reconstructions (Figure 3; see also Appendix 2). The methods employed here are not perfect; however, they provide an independent test of previous optical and X-radiograph observations and augment our knowledge of the morphology of this specimen. Comparison with light photographs demonstrates the superiority of CT techniques in eliminating lighting artifacts such as shadow and specularity, and provides a clearer illustration of topology and true morphology (Appendix 2). Additionally, the protocols outlined retain digital data representing a facsimile of the specimen, and therefore a testable data set (Appendix 1). Moreover the methods described herein can be duplicated and improved upon if so desired and therefore presents a superior level of testability and reproduction of results compared to optical examination or X-radiographs alone.
Figure 3.1-3.2 display isosurface images of the dorsal and ventral surfaces of the skull of Pachyrhachis. The ventral surface is reversed so landmarks can be stacked in order to recognize and remove distortion (Figure 3.1, see also Appendix 3). A line from the center of the basioccipital through the parasphenoid rostrum defines the midline of the skull (Figure 3.1). The posterior midline of the dorsal surface indicated by the sagittal crest lies above the ventral midline (Figure 3.2). As the skull was crushed, the snout rotated to the left, with the right maxilla overlying the tip of the right dentary and the left dentary displaced laterally but lying on its medial surface. The skull of Pachyrhachis is crushed dorsoventrally, with individual elements suffering varying degrees of compaction and displacement. The left and right postorbitals are symmetrically displaced laterally, and the coronoid processes of the lower jaws are collapsed medially on both sides. The symmetry of structures and their crushing patterns across the skull are important because they indicate force applied orthogonal to the bedding plane on which the specimen was preserved.
To quantify the amount of displacement of each element, the ventral isosurface reconstruction was rotated 18.4 degrees relative to the original scanning axis to approximate the alignment of the center of the parasphenoid and the posterior center of the basioccipital (Figure 3.1) at a reference angle of 0.0 degrees. The dorsal isosurface was then rotated -18.4 degrees to match the alignment of the ventral surface. By comparing the dorsal sagittal alignment of the parietal to that of the basioccipital-basisphenoid, it is clear that the braincase behaved as a single unit with respect to crushing (Figure 3.2; see also Appendix 3).
The parietal-basicranium midline was used as a reference of 0.0 degrees. In dorsal view the posterior terminus of the medial suture of the frontals is displaced slightly right of center, the suture angles 15.5 degrees to the left anteriorly. The prefrontals approximate this displacement to an equivalent degree. The right maxilla is preserved in a slightly more anterior position than the left. The small edentulous premaxilla is preserved in place between the anterior maxillaries. The right mandible is in articulation with the distal quadrate and displaced anteriorly to the left 29.0 degrees, underlying the skull and invading the natural position of the right ectopterygoid. The left mandible is preserved adjacent to the left side of the skull, embracing the left maxilla.
All relative displacement of elements including the snout, frontals, prefrontals, the left mandible, the left supratemporal, and the sagittal crest is leftward. All rotation of elements is counterclockwise, including the quadrates, with the left displaced medially and slightly overlapping the anterior vertebral column.
The morphological model (Figure 4.1) was constructed by building polygon surfaced wireframe simulation elements using Lightwave 3D, version 8 (Newtek 2004) and employing the three-dimensional isosurface model derived from the CT data as a guide. Removal of distortion from the elements was somewhat subjective; however, care was taken to approximate surface areas and lengths of the CT model in the proxy models of individual elements. The proxy elements were then manipulated to determine best fit to one another. Finally, the elements were distorted to mimic the distortion present in the CT data representing the actual fossil by manipulating the models (Figure 4.2). The undistorted and distorted reconstructions were then used as end points in a time sequence animation to simulate the interaction of elements and the relative timing and effects of crushing (Figure 4.3-4.9; see also Appendix 4 for animation sequence).
HUJ, Hebrew University of Jerusalem; UCMP, University of California Museum of Paleontology; TMM, Texas Memorial Museum, Austin Texas; CAS, California Academy of Science.
Pachyrhachis problematicus HUJ-PAL 3659; Cylindrophis ruffus CAS 231481; Cylindrophis ruffus UCMP 136995; Anilius scytale TMM(VPL) M-8281; Xenopeltis unicolor TMM(VPL) M-8276; Xenopeltis unicolor TMM(VPL) M-8277; Python regius TMM(VPL) M-8278; Python curtis TMM(VPL)M-8279; Epicrates cenchria TMM (VPL) M-8280.