Correlation of 2D Reconstructed High Resolution CT Data of the Temporal Bone and Adjacent Structures to 3D Images Rodt T 1, Ratiu P 1, Becker H 2, Schmidt AM 2, Bartling S 2, O'Donnell L 3, Weber BP 2, Jolesz FA 1, Kikinis R 1 ; Surgical Planning Laboratory, Brigham and Women's Hospital, Harvard Medical School 1, Boston, MA; Medical School Hannover 2, Hannover, Germany; Massachusetts Institute of Technology 3, Boston, MA Presented as a Poster at the 39 th Annual Meeting of the American Society of Neuroradiology (ASNR) 4/21/2001-4/27/2001, Boston MA Purpose The temporal bone anatomy is difficult to comprehend by looking at cross-sectional CT-images alone. 3D-visualization can help to understand the spatial relationship of the bony parts of the skull base, nerve structures and the middle and inner ear (1,2,3). 2D-images, however, often reflect the anatomy more accurately as some detail is lost by post-processing (3,4). Correlation of 2D-images to 3D-models might help to combine the advantages of these two modalities. Materials & Methods Multi-Slice CT scans covering both temporal bones were obtained in 17 patients with 1.25 mm collimation (140 kv, 40 ma, pitch 3). Two normal and 15 temporal bones with different pathologies including malformation, trauma, implants, tumor and inflammatory disease were investigated. A bone reconstruction algorithm was applied (0.3 mm reconstruction interval, 9.6 cm FOV) (5). Post-processing was performed using the in-house software "3D-Slicer". Segmentation was performed using thresholding and manual segmentation. The temporal bone was segmented as well as the ossicles, the tensor tympani muscle, the inner ear structures, the cranial nerves VII and VIII, and additional relevant structures in the case of pathology (6). 3Dmodels of the individual structures were generated in different colors using a surface-rendering algorithm. The 2D-slices were correlated to the 3D-models of the entire region by means of orienting the models along with the cross-sectional images into standardized views. The segmented areas for each individual structure were displayed in different colors in the axial, coronal and sagital images.
Results 3D-visualization of all the different anatomical structures displayed the correct anatomy and pathology in all the cases. The software "3D-Slicer" allowed various ways of interaction with the 3D-model, enabling a structured presentation of the anatomy and pathology. Post-processing took up to six hours due to time-consuming manual segmentation, merging of segmentationvolumes and generating the models. With more advanced post-processing algorithms this likely to be improved. In all the cases diagnosis would have been possible with the 2D-images alone. But correlation of the cross-sectional image to the 3D-model helped to understand the location and orientation of the 2D-slice in the 3D-anatomy. The complex spatial relationship of the individual structures of the temporal bone was easier to understand in the 2D-image when 3D- information was added. Conclusion Correlation of cross-sectional images to 3D-models is a more effective way of reaching a spatial understanding of the individual anatomy. Diagnosis can thus be made easier and the method can further be used for surgical planning and medical education. References 1)Howard JD, Elster AD, May JS. Temporal Bone: Three-dimensional CT. Part I. Normal Anatomy, Techniques, and Limitations. Radiology 1990; 177:421-425 2)Reisser C, Schubert O, Forsting M, Sartor K. Anatomy of the temporal bone: detailed threedimensional display based on image data from high-resolution helical CT: a preliminary report. Am J Otol. 1996; 17:473-479 3)Schubert O, Sartor K, Forsting M, Reisser C. Three-dimensional computed display of otosurgical operation sites by spiral CT. Neuroradiology 1996; 38:663-668 4)Howard JD, Elster AD, May JS. Temporal Bone: Three-dimensional CT. Part II. Pathologic Alterations. Radiology 1990; 177:427-430 5)Casselman JW. Temporal bone imaging. Neuroimaging Clin N Am. 1996; 6:265-289 6)Janecka IP, Tiedemann K. Skull Base Surgery; Anatomy, Biology and Technology, Lippincott-Raven Publishers 1997
Fig.1: Generating a 3D-model and correlation to the 2D-image: 1A: reconstructed axial CT image. 1B: segmentation of the anatomical structures. 1C: 3D-model is generated based on segmentation and correlated to axial 2D-image. 1D: superior view of 3D-model of middle and inner ear structures correlated to the axial 2D-slice.
Fig.2: Tumor of the temporal bone 2A: reconstructed axial CT shows bone erosion, the overall amount of erosion is difficult to evaluate. 2B: lateral view of correlated 2D-image and 3D-model. The ossicular chain is still intact. 2C: lateral view of 3D-model shows amount of bone erosion but spatial relation to 2Dimage is difficult to comprehend. Fig.3: Total Ossicular Replacement: 3A: reconstructed axial CT shows tip of prosthesis posterior to oval window. 3B: anterior view of correlated 2D-image and 3D-model displays spatial relationship of inner ear structures and 2D-image. 3C: 3D-model shows discontinuity but spatial relation to 2D-image is difficult to comprehend. Fig.4: Fracture of the temporal bone and dislocation 4A: axial CT shows dislocation of the incudomallear joint and fracture fragment. 4B: inferior view of correlated 2D-image and 3D-model shows dislocation of the ossicular chain caused by impingement of the fracture fragment. The fracture course is shown. 4C: inferior view of 3D-
model shows pathology but spatial relation to 2D-image is difficult to comprehend.