Roger A. Horton

Comparison of film, direct digital, and tuned-aperture computed tomography images to identify the location of crestal defects around endosseous titanium implants

This study compared diagnostic performance obtained from two-dimensional and three-dimensional x-ray images. The latter were produced with a new tomosynthetic method based on aperture theory called tuned-aperture computed tomography. Seven human cadaver mandibular segments containing a total of 20 endosseous implants with a small randomly positioned alveolar crestal defect at each implant site were imaged in two dimensions with periapical film and with a charge-coupled digital detector, and digitally with the same detector in three dimensions with tuned-aperture computed tomography and subtracted tuned-aperture computed tomography techniques. Seven trained dentists viewed randomized displays of all modalities. Outcomes of the diagnostic task of identifying the locations of crestal defects were quantified with accuracy, confidence, and time performance measures. Analyses of variance demonstrated that differences between either three-dimensional technique and either two-dimensional modality were significant for all measures (p < 0.001). These findings suggest that clinically applied TACT methods hold promise as an improvement over the status quo.

TACT Imaging of primary caries

OBJECTIVE Tuned-aperture computed tomography, a new method for creating 3-D radiographic information based on optical aperture theory, was evaluated for diagnostic efficacy in primary caries detection. STUDY DESIGN Sixty-four extracted teeth with 89 carious lesions were imaged with D-speed film, direct digital, and TACT modalities. A commercially available, 8-bit, charge-coupled device was used in the later two modalities. Six trained observers were asked to identify the presence or absence and depth of interproximal and occlusal lesions for all three modalities. The teeth were sectioned and examined microscopically to determine ground truth. Logistic regression analysis was performed for all three imaging systems for the detection task. Analysis of variance was used for depth determination. Detection of lesion, depth of lesion accuracy, and time for diagnosis were also examined. RESULTS For caries detection TACT and film were not different (p = 0.2216) with the Wald statistic. Film and TACT were significantly more accurate than the digital system (p = 0.0001). Scheffes post hoc test revealed that TACT and film were more accurate than the direct digital system for determining lesion depth (p = 0.05) but not statistically different when compared with each other. The detection data were substantiated further by receiver operating characteristic analysis that demonstrated similar statistical relationships. Time required per diagnosis was not shown to be statistically different among the three imaging modalities. CONCLUSIONS We conclude for caries detection and depth determination that TACT could not be distinguished from film despite the significant relative loss of information capacity in the charge-coupled device receptor. The relatively poorer performance yielded by the digital control images suggests that increased information capacity associated with more modern charge-coupled device detectors may improve diagnostic performance for both direct digital and TACT displays over that demonstrated in this investigation.

Detection of peri-implant bone changes with axial tomosynthesis.

Tuned aperture computed tomography was used to assess bone defects at implant sites that are normally obscured in conventional periapical projections. Titanium implants were placed in incisor, premolar, and molar areas of an edentulous dry human mandible. The alveolar crest adjacent to each implant was circumferentially grooved with progressively larger round burs (no. 1/4 to no. 6). After each increment of bone removal, axial tuned aperture computed tomography slice images were produced. From these, digital subtraction images were generated with incremental and baseline slices. A panel of eight dentists reviewed randomized pairs of images that showed defects that differed by a single bur size. Images were masked to limit observation to mesial, distal, facial, or lingual regions. Observers accurately identified larger lesions in 76% of the unsubtracted tuned aperture computed tomography images and in 90% of the subtracted images. Logistic regression analysis suggested significant differences as a result of lesion size (p < 0.036) and imaging modality (p < 0.020).

Predicting osseous changes in ankle fractures

A quantitative assay of subtle morphologic changes that are induced in the trabecular structure of weight-bearing bones as a result of immobilization prescribed as part of routine clinical management is investigated. It is postulated that a clinically meaningful structural measure can be inferred from comparatively small, relatively uniform regions of trabecular bone, which are easily isolated from conventional transmission radiographs. This approach derives from the observation that common transmission radiographs of alveolar bone contain trabecular regions with spatial attributes that are statistically independent of image magnification over a reasonable range of spatial frequencies. Selecting these regions for analysis is advantageous because certain attributes of the resulting data are theoretically independent of the focal-object distance used to produce the original radiograph. Therefore, the size and shape of the irradiated tissue mass do not influence the measurement, as they do in morphometric-based analysis.<<ETX>>

Effects of projection geometry and number of projections on accuracy of depth discrimination with tuned-aperture computed tomography in dentistry

OBJECTIVE The purpose of this study was to determine the degree to which the number and angular disparity of component projections influence depth discrimination with tuned-aperture computed tomography. STUDY DESIGN Groups of three tiny steel spheres served as fiducial references on and in four partially edentulous mandibles. Two spheres were attached to the facial and lingual surfaces of each mandible, and the third was fixed in the apical region of an open tooth socket. Errors in estimates of the depth of the apically positioned sphere relative to the other two spheres were determined from three-dimensional tuned-aperture computed tomography reconstructions. These data were compared with actual measurements produced independently with an optical micrometer. Multiple projections required by the tuned-aperture computed tomography reconstruction algorithm were produced from radially symmetric exposures bearing angular disparities of 5, 15, 30, and 45 degrees. The number of symmetrically dispersed projections per tuned-aperture computed tomography reconstruction likewise was varied systematically (2, 4, 8, 12, and 16 projections). These variables were manipulated through the use of a balanced factorial design. Depth estimates were performed by trained observers; the estimates were based on the determination of tuned-aperture computed tomography slices perceived as imaging the respective apical spheres in sharpest focus. Specimen and observer effects were also considered as independent variables. Resulting data were normalized by logarithmic transformation and analyzed statistically by analysis of variance. RESULTS Significant differences (p < 0.005) were demonstrated for angular disparity and specimen effects, but the number of projections and the effect of the observer were not found to be statistically significant. CONCLUSIONS In dentistry, angular disparities of 15 degrees or greater should be used when tuned-aperture computed tomography is being applied to diagnostic tasks requiring maximal depth discrimination accuracy.

Effect of number of projections on accuracy of depth discrimination using tuned-aperture computed tomography for 3-dimensional dentoalveolar imaging of low-contrast details

OBJECTIVE The purpose of this study was to test the hypothesis that the number of projections influences the accuracy of a simple depth discrimination task when tuned-aperture computed tomography is used. STUDY DESIGN In each of 4 partially edentulous mandibles, 2 radiopaque steel spheres were attached to the facial and lingual surfaces and 1 ceramic sphere was place in the apical region of an open tooth socket. Errors in estimates of the depth of the apically positioned ceramic sphere relative to the 2 steel spheres were determined from 3-dimensional tuned-aperture computed tomography reconstructions. These data were compared with actual measurements produced independently by means of an optical micrometer. Multiple projections were produced from radially symmetric exposures bearing an angular disparity of 15 degrees. The number of symmetrically dispersed projections per tuned-aperture computed tomography reconstruction was varied systematically (2, 4, 8, 12, and 16 projections). The consequences of this variable, as well as specimen and observer effects, were evaluated in a balanced factorial experimental design. Depth estimates were performed by 10 trained observers. The depth reported was that corresponding to the tuned-aperture computed tomography slice perceived to yield the image of the ceramic sphere in sharpest focus. Resulting data were normalized by logarithmic transformation and analyzed statistically by analysis of variance. RESULTS No statistically meaningful effects were found for the number of projections (P = .607) or for different observers (P = .093), but a significant specimen effect was demonstrated (P = .006). CONCLUSIONS Factors other than high contrast limit the perception of image sharpness under these conditions. Depth may be estimated accurately from relatively small numbers of projections.

Use of tuned aperture computed tomography (TACT) with a large-scale imaging panel for 3D tomography of the skull

Surgeons often need to localize foreign objects in wounded patients in facilities where computed tomography (CT) may not be present or is contraindicated by anticipation of massive CT artifacts associated with imaging radiopaque materials such as shrapnel. A skull phantom was placed on top of a 22 by 18 cm radiographic imaging panel developed by Du Pont Diagnostic Imaging Systems in both posterior-anterior (PA) and lateral orientations. A radiopaque reference sphere (2 mm dia.) was attached to the phantom and a set of eight, two- dimensional radiographic projections was obtained for each of the two orientations. This was done by laterally displacing the x-ray head in a plane parallel to the projection panel to eight different positions well distributed around the periphery of a solid angle approximating 25 degrees. An operator identified the reference shadow on each digital projection. This information was used by the TACT system to compute tomographic slices based on the reference spheres relative displacement in the field. These TACT slices were compared directly with tomographs of the same region of the skull made with a Phillips Model BTS4 tomographic machine. Both methods appeared to yield clinically interpretable images of comparable diagnostic quality.