1996 Volume 2 Issue 2 Pages 91-100
In order to develop an automatic contour extraction procedure of the cortical bone surfaces of the condyle and glenoid fossa of temporomandibular joint (TMJ) from Magnetic Resonance Images (MRI), a series of image processing techniques were applied to TMJ-MRI and the resulting images were evaluated. A TMJ in a healthy male was imaged on a 0.5 T system MRVectra (GE Yokogawa Medical System, Tokyo), with a 9 cm diameter surface radio frequency receiver coil in sagittal plane. A spoiled GRASS (SPGR) sequence with a 51 msec repetition time (TR) and a 12 msec echo time (TE) was used to obtain a series of 30 contiguous, 1 mm sections in sagittal plane. With a field of view of 6.5 cm by 13 cm and a resolution of 128 by 256, pixel sizes of 0.5mm by 0.5mm were measured and provided for image processing. Region segmentation using thresholding was employed for extracting the region of the condyle and fossa consisting of the cortical bone. As a threshold selection method, mean adjacent-pixel number method was applied in each slice, because the strength of the MR signal in each of serial slices were different due to the distance from the surface coil. This method provided appropriate threshold values to obtain adequate binary images of the cortical bone region, irrespective of the proportion of cortical bone areas in the slices. However, the binary images, in which the condyle and fossa formed a connected component, were observed because of low signal to noise ratio of the original images. Here intraarticular space between the condyle and fossa, with a width of at least 1 pixel, was presumed based on the anatomical structure of TMJ. Thus, a binary image of the presumed intraarticular space was obtained using thresholding and thinning on the image treated with low pass filtering technique and utilized to lose connectivity of the condyle and fossa. Smoothing for binary images was then performed to extract the contour of the condyle and fossa, respectively. The region of the intraarticular space derived from the slice was processed with thinning and used to disconnect the condyle and fossa in the adjacent slices. Then the smoothing for the contour extraction was repeated in the adjacent slices. Resulting images of the contour of the condyle and fossa in each slice well coincided with the manual traces of those structures by experienced dentists. The resulting 2-D images were reconstructed to 3-D images by means of 3-D visualization program (AVS Inc., USA) on graphic workstation. The reconstructed 3-D images represented the anatomical characteristics of TMJ region. These results indicate that the contour extraction method investigated in this study was appropriate to extract the cortical bone surfaces of TMJ structure from MRI.
