AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 821283-294 (1990)
Validation Study of Skull Three-Dimensional Computerized Tomography Measurements CHARLES F. HILDEBOLT, MICHAEL W. VANNIER, AND ROBERT H. KNAPP Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, Missouri 63110
KEY WORDS
Craniometry, Anthropometric, 3-D CT, CT
ABSTRACT
Recent advances in imaging have led to high-resolution computerized tomography (CT) scanning with exquisitely detailed slice images of the skull and three-dimensional (3-D) surface reconstructions using computer software. It is possible to use CT scans to acquire morphologic information about the skull in a convenient digital form and to derive 3-D measurements from surface reconstruction images. Unfortunately, no effort has been made to date to test the validity of these measurements on laboratory specimens, and no compelling evidence is available from phantom studies to indicate the nature and magnitude of the errors inherent in the measurement technique. We have performed a pilot study to quantify the morphology of the skull based on surface features that can be found in CT scans and 3-D reconstructions. Comparative measurements were obtained from five skulls (two normal and three with dysmorphology) with calipers and a 3-D electromagnetic digitizer. These measurements were statistically compared with those based on original CT scan slices and reformatted 3-D images. It is concluded that 3D-CT measurement techniques are superior to those in which measurements are obtained directly from the original CT slices; 3-D CT methods, however, must be significantly improved before measurements based on these techniques can be used in studies that require a high degree of precision. The results are used to indicate the most fruitful areas of future study.
In comparative anatomical studies, major concerns are 1) to quantify differences in forms and 2) to quantif form changes through time. Such quanti ication should be accurate, reproducible, and time efficient. Traditionally, quantification of skull form has been based on measurements made with simple rulers, calipers, and goniometers. The major limitations of these measurin devices are that they are time intensive an only surface measurements of external features are possible. Cephalometrics overcomes one of these limitations in that both internal and external structures (Broadbent et al., 1975) can be visualized; however, cephalometric radiographs are limited due to superimposition of structures, loss of skin surface detail due to X-ray transparency of soft tissues, and significant geometric magnification error due to the divergent nature of the X-ray beam.
P
3
@ 1990 WILEY-LISS, INC.
In the mid-l970s, computerized tomography (CT) scanning became available. Today several thousand CT scanners are in operation throughout the world. CT scans overcome man of the shortcomin s of conventional sku1 X-rays (Newton an Potts, 1981). They minimize superimposition of structures, possess high soft tissue contrast, and have essentially no geometric magnification error. A complete high-resolution CT scan examination of the head may, however, require more than 100 slices of 2 mm thickness each. It is difficult to work with such a large number of views and identify s ecific landmarks. Moreover, very few eop e are able to combine the slices menta ly into the 3-D images necessary to the evaluation of anatomical structure.
i
f
P f
Received June 20,1988; accepted October 16,1988.
284
C.F. HILDEBOLT ET AL.
To overcome these limitations, software programs were developed that transform CT scan slice data into a familiar and readily comprehensible 3-D format (Vannier et al., 1983a,b; 1984a,b; Marsh and Vannier, 1985). Methods and software were further developed to allow 1)the removal of overlyin anatomic structures, revealing deeper hi den structures and 2) the display of hard and soft tissue from any desired viewing perspective, including unobstructed views of the endo- and exocranial bases (Vannier et al., 1983a, 1984a,b, 1985; Dye and Vannier, 1984; Gayou et al., 1984; Marsh et al., 1984, 1985a,b, 1986; Stevens et al., 1984; Totty and Vannier, 1984; Knapp and Vannier, 1985; 1986; Weeks et al., 1985; Marsh and Vannier, 1987). Moreover, over 50 software packages have been develo ed for performing 3-D reconstructions o CT slices (Huijsmans et al., 1986). At many institutions, the use of 3-D CT reconstructions has become routine in qualitative evaluation of abnormalities of the head and in determinations of treatment protocols. There is growing speculation that 3-D CT may be particularly applicable to anthropometric and craniometric studies of prehistoric (Conroy and Vannier, 1984; Conroy and Vannier, 19871, living-, and deceased-contemporary populations, including Australasian populations. To date, however, no one has demonstrated the quantitative value of 3-D CT anthropometric measurements. This limitation is not readily ascertainable from the glowing descriptions often gwen in professional journals. It is clear that many people consider 3-D CT ima ‘ng to be a quantum advancement, but is t is the case? The ultimate goal in producing 3-D images is to convey information about objects to observers (Hanson, 1985; McCormick et al., 1987). The value of this information depends on its usefulness. A great deal of time, money, and thought have gone into using advanced technology to produce 3-D CT images. So far, most of the 3-D CT effort has been directed toward making the “pretty pictures” prettier. Terms such as “fidelity,” “ recision,” “quality,” and “artifact” have een used as artistic qualifiers rather than as scientific descriptors. Because the terms have not been defined with regard to 3-D CT and because statements involving these terms have not been supported with adequate quantitative data, evaluations of image-processing parameters
3
P
f
B
and judgments of image usefulness are of limited value. Moreover, without quantitation, one cannot com letely characterize norma1 and abnormal s ull morphology or measure the effects of growth or treatment. The study that we report is an initial step in determining the anatomical fidelity of 3-D CT images.
I:
MATERIALS AND METHODS
The objective of this study was to com are measurements based on original CT s ices and 3-D reconstructions with measurements made with calipers and validated with a 3-D digitizer system. For our study, we used the mean of original and repeat caliper measurements as truth and refer to the agreement between truth and CT-slice and 3-D CT measurements as “precision.” Standard spreading and sliding calipers were used to make the craniometric measurements, as given in commonly used osteological texts (Bass, 1981; Brothwell, 1981). The 3-D digitizer system (Hildebolt and Vannier, 1988) is an electromagnetic six-degree-of-freedom digitizer (3Space Digitizer/Tracker, McDonnell Douglas Electronics Company, Polhemus Navigation System Division, Colchester, VT) connected to a personal computer (Macintosh Plus, Apple Computer Corporation, Cupertino, CA) that was interfaced to a CAD/ CAM system (Unigraphics computer-aided design and computer-aided manufacturing system, McDonnell Douglas Manufacturing Systems Corporation, Cypress, CAI. Analyses were limited t o the measurements of bony anatomical landmarks of dry skulls. Landmarks and measurements are listed and defined in Table 1and are illustrated in Hildebolt and Vannier (1988).Additional information on landmarks is contained in Bass (1981) and Brothwell (1981). Five adult human skulls were selected for measurement. Two of the skulls were normal. The other three were from the museum collection of Dr. Paul Tessier (Paris, France) and represent the abnormalities of unicoronal synostosis, orbital neurofibromatosis, and nasofrontal meningoencephalocoele.The precision of the 3-D digitizer measurements compared well with caliper measurements, there being no significant difference (P < 0.05) between measuring techniques (Hildebolt and Vannier, 1988). CT scanning was performed with a standard clinical instrument with third-generation geometry (Somatom DR-H, Siemens
P
3-D SKULL MEASUREMENTS
285
TABLE 1. Landmarks and measurements used in study Landmarks A As B Br
D Enm Eu G Gn
I La Mn N Ns 0 Or OP PCP S V Zm ZY Measurements As-As BiW B-A B-Rr Br-N B-0 B-N B-Pcp B-V Br-La D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A NaBrd N-Gn N-Pcp Ns-N OrBrd Or-S zy-zy Zm-Zm
Alveolare, tip of maxillary alvolar bone between central incisors Asterion, junction of occipital, parietal, and temporal bones Basion, anterior midpoint of foramen magnum Bregma, junction of the sagittal and coronal sutures Dacryon, point at which frontal, maxillary, and lacrimal bones meet Endomolare, medial point of alveolar ridge a t the second molar The two Eu points establish a line that represents the greatest width of the skull Glabella, most anterior part of skull in midsagittal plane Gnathion, lowest point on mandible Infradentale, tip of mandibular alveolar bone between the incisors Lambda, junction of sagittal and lambdoid sutures; Mentale, most anterior point of mental foramen Nasion Nasospinale, highest point on nasal spine Opisthion, posterior midpoint of foramen magnum Orale, midpoint between central incisors on a line tangent to their posterior surfaces Opisthocranium, most posterior part of skull in midsagittal plane Posterior clinoid process Staphylion, point where a line tangent to the two curves a t the posterior of the palate crosses Vertex, highest point of skull Zygomaxillare, lowest point on the suture between the zygomatic and maxillary bones Zygion, most lateral point of zygomatic arch Biasterionic breadth Bicondylar width, distance between mandibular condyles. Basi-alveolar length Basi-bregmatic height Frontal chord Foraminal length Basinasal length Basiposterior-clinoid-process length Skull height Parietal chord dacryonic chord palatal breadth Foraminal breadth, maximum width of the foramen magnum. Skull length Symphyseal height Occipital chord foramen mentalia breadth Upper facial height Nasal breadth, maximum width of nasal opening perpendicular to sagittal plane Total facial height Nasio-posterior-clinoid-process length Nasal height Orbital breadth, maximum width of orbit measured from D Palatal length bizygomatic breadth maxillary breadth (midfacial breadth)
Medical Systems, Inc., Iselin, NJ). Two-millimeter, axial (perpendicular t o the long axis of the body), contiguous slices were made every 2 mm. These slices were parallel to the Frankfort horizontal plane, which passes through the most inferior margins of the bony orbits and the most superior margins of the auditory meatuses. Level-slicingcontour extraction was used to perform 3-D surface reconstructions. An observer (C.F.H.) recorded the X, Y, and Z coordinates of skull landmarks from CT slices, using the LI
(line), CO (column), and TP (table position), respectively (Fig. 1).The distances between landmarks were computed, with appropriate magnification correction factors being applied to the X and Y coordinates (when picture magnification was used). Measurements between landmarks were made directly from the 3-D CT reconstructions (Fig. 2) by means of an interactive computerized CT scan evaluation console (Siemens Evaluscope DR). Coordinates of landmarks, magnification factors, and 3-D CT distances
Fig. 1. Determination of coordinates of a point (see arrow on CT slice). The LI (line), CO (column), and TP (table position) represent, respectively, the X, Y, and Z coordinates of the point. TP represents the position ofthe skull in the scanner; that is, TP 1represents 1mm, and,
in this case, TP 79 represents the 79 mm from the starting position set by the operator on the CT scanner table. The LI and CO values are determined automatically with software that operates on the evaluation console.
Fig. 2. Illustration of dacryonic cord measurement on a 3-D CTreconstruction (see 1on image).The distance between landmarks (in this case, dacryons) is computed
automatically with software operating on the evaluation console.
3-D SKULL ME ASUREMENTS
287
were all determined with software that oper- several areas of the unicoronal synostosis, ated on a stand-alone CT scan evaluation orbital neurofibromatosis, and one of the console. All CT coordinate measurements normal skulls were not available for analysis were repeated over a 6 month period. from the 3-D CT reconstructions. Seven (5%) The percentage of variation between sets of the measurements, therefore, could not be of measurements (original vs. repeat) and made. Of the remaining 114 measurements between measurement techniques (caliper that could be made from 3-D CT reconstrucvs. CT slices and caliper vs. 3-D CT recon- tions, 27 (24%)could not be performed bestructions) were determined with the follow- cause landmarks could not be confidently ing formula. identified (Table 3). Results of CT-slice measurements and 3-D CT measurements are given in Tables 2 and 3. Standard deviations, N I minima, maxima, means of differences, and percent of incongruence between measurements (original vs. repeat, caliper vs. CT slices, and caliper vs. 3-D CT)were tabulated (Tables 4-7). The largest “mean difference” between where Xi, is a measurement in the firstset, ori ‘nal and repeat measurements made Xi, is a measurement in thesecond set, Xi, is wit TI CT slices was for FoBrd (2.2 mm; meanvalue for the first set, Xi, is mean value for the second set, and N is the number of Table 4). This is a 7.24% incongruence. Seven additional mean differences were measurements. The reliability of measurements (original 1mm or more. The largest “maximum differvs. repeat, caliper vs. CT slices, and caliper ence” between original and repeat CT-slice vs. 3-D CT) was assessed by calculating the measurements was 6.8 mm (FoBrd), with 11 means of the differences (between measures of the maximum differences exceeding2 mm. collected by the various methods) and the In the comparison of CT-slice and caliper standard deviations of the means of the dif- measurements, the largest mean difference 5.6 mm (Or-S; Table 51, which is a ferences (that is, we compared two sets of was 12.11% incongruence. Fourteen of the remeasures taken with the same instrument and then compared the mean of these mea- maining mean differences exceeded 2 mm. sures to similar means for other instru- The largest maximum difference was 9.2 mm ments). Student’s t tests and paired t tests (Zm-Zm), with 16 of the remaining maxidifferences exceeding 3 mm. (SAS Institute Inc., 1985) were used to test mum The largest mean differences between the null hypotheses that there were no diforiginal and repeat 3-D CT measurements ferences between the means of original vs. were for Or-S and As-As (3.5 mm; Table 6). repeat measurements and caliper vs. other These represented incongruences of 7.33% forms of measurement. The correlation coefficients for these measurements were also and 3.33%, respectively. Fourteen of the redetermined, as were the standard errors for maining mean differences equalled or exmodels with forced intercepts of zero and ceeded 1 mm. The largest maximum differslopes of one. For comparing techniques, the ence was 7.0 (As-As), with 17 of the averages of the two measurements (original maximum differences equalling or exceeding 2 mm. For variations between caliper and and repeat) were used. 3-D CT measurements, the largest mean difference was 3.4 mm (Mn-Mn;Table 7), for RESULTS a percent incongruence of 7.66. Four of the The top of the calvaria was missing from remaining mean differences exceeded 2 mm. the nasofrontal meningoencephalocoele The largest maximum difference was skull; so Br, V, and La could not be mea- 5.5 mm, with eight of the remaining maxisured. The calvariae of the unicoronal synos- mum differences exceeding 3 mm. tosis skull and one of the normal skulls were The results of Student’s t and paired t tests intact and Pcp could not be obtained with indicate that none of the differences between caliper measurements. Thus gold-standard the means of measurements were significant caliper determinations were not available (P > 0.05) except for Br-N for CT-slice vs. for nine (7%) of the measurements. Because caliper measurements. All the regressions of software and field of view limitations, were significant (P < 0.001) with correlaI
,
288
C.F. HILDEBOLT ET AL. TABLE 2. Original and repeat skull measurements f r o m CT slices
Measurement
Normal 1 (mm) 115 113 132 116 03: 132 121 124 109 026 034 037 204 033 099 048 070 121 030 054 051 041 089 136
As-As R-A B-Br B-N
B-0 B-Pcp Ba-V BiW' Br-La Br-N D-D Enm-EnLm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N Or-S OrBrd Zm-Zm Zv-zv
116 113 133 117 032 132 121 124 109 025 037 037 205 031 098 049 070 12J 030 055 051 039 089 136
Unicoronal synostosis (mm) 098 096 124 100 03z -
119 111 101 108 022 034 028 170 029 094 040 070
"9,024 052 042 037 098 131
Skull Meningoencephalocoele (mm) 094 093 08f! - 08; -
098 096 125 101
102 101 038 038 039 04t - -
03! 119 111 102 108 021 035 027 171 029 092 040 071 113
11:
-
024 052 045 038 097 131
11 : -
-L
-2
028 033 029 177 035 -
031 034 028 177 032 -
050 078 126 072 025 065 041 033 098 129
05 1 077 126 072 025 065 041 033 098 129
Orbital neurofibromatosis (mm) 105 092 128 095 036 036 123 111 125 110
016 031 029 176 022 084 043 059 099 075 025 046 040 033 078 127
105 096 127 096 034 035 121 108 125 110 015 031 026 178 023 086 043 060 099 074 025 044 039 044 078 126
Normal 2 (mm) 093 093 130 102 033 043 131 111 108 109 179 025 031 176 029 093 038 065 111 07 1 022 044 046 039 078 119
094 094 131 102 034 042 132 112 108 109 179 025 026 170 029 092 038 067 111 072 023 044 045 039 078 119
'Bicondylar width. > l o p of calvaria missing. "Calvaria intack; therefore, gold-standard mpasurrments not pvssihle.
2.
tions greater than 0.994 (the standard error for caliper vs. 3-D CT was 0.042 and for caliper vs. CT slices was 0.099). DISCUSSION
In general, we found no statistically significant differences between the measurement techniques under investigation; however, given the small sample sizes of our pilot study, type I1 errors may have been committed (that is, differences may actually exist). An expanded study is needed to clarify this point fully. At any rate, we suggest that there are substantive differences between the techniques, for the following reasons. Ideally measurements based on CT slices and 3-D CT reconstructions should be as precise as measurements made with calipers. In other words, the difference between caliper and CT-slice or 3-D CT measurements should not be any greater than are repeat Cali er measurements. In an earlier study, we etermined the amount of variation between original and repeat caliper measurements for the same 26 measure-
B
ments used in this study (Hildebolt and Vannier, 1988). In that study, the mean difference between original and repeat caliper measurements exceeded 1 mm for only one measurement. In our current study, when comparing caliper with CT-slice measurements, only four of the mean differences were less than 1mm, and the same was true for 3-D CT. Moreover, for CT-slice measurements, 15 of the mean measurements exceeded 2 mm, seven exceeded 3 mm, three exceeded 4 mm, and one exceeded 5 mm. 3-D CT measurements were considerably better, with only five measurements exceeding 2 mm and one measurement exceeding 3 mm. The superiority of 3-D CT to CT-slice measurements is also generally supported by the results presented above. We suggest, however, that there is considerable room for improvement with regard to the precision of measurements made from both CT slices and 3-D CT reconstructions. This is discussed below. There are problems with both techniques with regard to the time commitment re-
289
3-D SKULL MEASUREMENTS TABLE 3. Original and repeat skull measurements from 3 - 0 CT reconstructions
Measurement
Normal 1 (mmi
As-As R-A R-Hr R-N B-0 B-Pcp Ra-V BiW' Br-La Br-N
-2 111, 119 037 138 12if -
D-D
025 035 036 204 032 052 072 124,I 029 056 053 038 095 136
Enm-Enm FoBrd
G-Op I-Gn La-0 Mn-Mn N-A N-Gn
N-Pcp
NaBrd Ns-N Or-S OrHrd Zm-Zm Zy-zy
Unicoronal synostosis (mm)
-2 11; 116 03f 136 12f -
110 094 -5 099 036 036 -5 114
-J
-2 -2 027 038 037 201
04: 049 071 123 ,I 028 055 048 037 096 136
Skull Meningoencephalocoele (mm)
-5
027 028 169 169 03; 022 -
-
043 041 067 068 11: 11: 026 042 044 096 130
026 04,7 043 097 130
(mm)
-.I
-5
024 033 029 176 030
049 083 129 071 024 066 046 039
051 083 127 070 024 065 044 03J
-
Normal 2
107 093 -5 100 036 035 -5 114
025 033 029 177 029
-5
Orbital neurofibromatosis (mm)
-5
-
127 125
'Ricondylar width. 'Landmarks fnr measurement could not hc loreted in 3-D C T reconstruction. "Calvaria intack; therefore, gold-standard measurements not possible. :Missing due to field of view limitations. "Top of calvaria missing.
quired for optimal results. For CT-slice measurements, it is difficult to work with the large number of CT slices used in a scan of the skull (normal skull 1required 113 slices). It requires considerable time to look at each slice, and most people find it difficult to construct 3-D images of the skulls from memory. Such mental images are, however, necessary for one to be able to locate landmarks. Without exception, it is necessary to scroll back and forth among at least several slices to locate a landmark. In some instances (e.g., FoBrd, OrBrd, NaBrd, or EnmEnm), no amount of scrolling back and forth eliminates the ambiguity of identifying landmarks; that is, it is not possible to identify exactly which slice contains the desired landmark. With these limitations in mind, it is much easier to make measurements from 3-D CT reconstructions than from the original slices. Most landmarks, if visible on the reconstruction, can be located as quickly as they can on a dry skull, and the measurements can be
made even more quickly than with calipers. There are, however, two problems associated with locating landmarks on 3-D CT reconstructions. First, some points cannot be seen. For example, if the coronal and sagittal sutures are not patent, bregma cannot be seen. Suture enhancement, a prereconstruction processing procedure used to improve suture visualization, can be used to overcome this problem (Fig. 31, but this procedure requires that an operator edit all the CT slices that contain the involved sutures (Marsh et al., 1985a). This time commitment negates the time savings made by using 3-D CT reconstructions. It is also possible to enhance the sutures with gradient shading programs (Fig. 41, and in the future measurements will be made from these shaded images. The second problem with the 3-D CT reconstructions concerns locating landmarks on the computer display screen. If the X and Y axes describe the plane of the screen, and the Z axis is orthogonal (perpendicular) to the X and Y axes, then it is reasonably easy
290
C.F. HILDEBOLT ET AL TABLE 4. CT slice measurement variation Mean
Measurement
N
As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW' Br-La Br-N D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-PCD NaB;d Ns-N Or-S OrBrd Zm-Zm zv-zv
5 5 4 5 5 3 4 5 4 4
5 5 6 5 5 4 5 5 5 3 5 5 5 5 5 5
SD of A
Minimum
A' (mm)
(mm)
(mm)
0.8 1.0 0.7 0.9 1.0 0.9 0.6 0.8 0.5 0.3 1.2 1.2 2.2 1.9 0.6 1.3 0.4 0.9 0.2 0.7 0.7 0.4 1.o 0.6 0.2 0.3
0.8 1.7 0.3 0.6 1.2 0.3 0.9 1.0 0.5 0.2 1.1 1.1 2.8 2.4 0.9 I .0 0.4 0.9 1.3 0.8 0.8 0.7 1.2 0.5 0.4 0.4
0.0 0.0 0.4 0.0 0.0 0.5 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.2 0.0 0.1 0.0 0.0 0.0 0.0
A-A
0.1 0.0 0.0 0.0
Maximum (mm)
Percent incongruence
1.6 4.0 1.0 1.5 2.9 1.2 1.9 2.4 1.0 0.6 2.9 3.0 6.8 6.1 2.1 2.3 0.9 2.2 0.5 1.7 1.6 1.7 3.1 1.3 0.8 0.8
0.81 0.99 0.57 0.85 2.85 2.22 0.46 0.70 0.40 0.27 5.19 3.83 7.24 1.06 2.05 1.42 0.82 1.34 0.21 1.10 2.57 0.80 2.32 1.44 0.27 0.22
:Difference. Ricondylar width
TABLE 5. Variation between caliper and CT-slice measurements Mean
SD of A (mm)
Minimum A (mm)
Maximum
A'
(mm)
Percent incongruence
3.1 2.8 2.1 1.0 0.9 0.8 1.8 0.4 3.6 0.5 1.7 1.7 0.4 2.1 2.8 3.2 0.7 1.2 1.6 1.0 0.6 1.6 2.4 1.8 3.5 0.3
0.1 0.3 0.2 0.2 0.1 0.3 1.0 0.8 0.3 2.0 0.1 0.0 0.0 0.3 1.0 0.5 1.4 0.5 2.3 0.2 0.3 0.5 1.7 1.5 0.2 0.7
7.3 7.2 5.1 2.5 2.6 1.7 5.3 1.9 8.0 3.1 4.3 3.7 1.2 5.9 7.3 7.0 2.8 3.3 6.4 1.9 1.9 4.9 8.1 5.9 9.2 1.4
3.29 2.44 1.69 0.88 4.17 1.99 2.62 1.18 2.27 2.40 8.52 5.53 2.33 1.36 113 2 2.48 4.95 2.32 3.57 1.74 3.62 4.66 12.11 7.84 5.03 0.86
Measurement
N
(mm)
As-As B-A B-Br B-N B-0 B-Pcp Ba-V
5 5 4 5 5 ;3 5 5 4 4 5 5 5 5 5 4 5 5 5 3 5 5 5 5 5 5
3.4 2.4 2.2 0.9 1.5 0.8 3.3 1.3 2.6 2.6 1.8 1.8 0.7 2.4 3.4 2.3 2.1 1.6 4.2 1.3 0.9 2.4 5.6 3.0 4.5
RiW2 Br-La Br-N D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N
Or-S OrBrd Zm-Zm Zy-zy :Difference. Bicondylar width
1.1
A
29 1
3-DSKULL MEASUREMENTS TABLE 6. 3 - 0 C T measurement variation
Mean
SD
Minimum
A'
A
A
(mm)
(mm)
Percent incongruence
-
7.0 1.o
3.33 l.0,2
-
-
3.0 2.0 1.0 2.0 2.0
1.38 2.26 1.77 76 89
2.0 3.0 1.0 3.0 4.0 1.0 3.0 1.o 2.0 3.0 2.0 2.0 5.0 4.0
4.93 2.96 2.02 0.44 5.36 1.18 4.40 1.04 1.26 2.32 2.41 2.83 7.33 4.56 1.04 0.47
Measurement
N
(mm)
of A (mm)
As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW Br-La Br-N D-D Enm-Enm FoBrd G-OP I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N Or-S OrBrd Zm-Zm zv-Zv
2 4
3.5 1.o
5.0
0
0.0,
1.0,
0
-
5
1.4 0.8 0.7 1.0
0.9 0.8 0.6 1.4
1!.
0.7,
1.0 0.0 0.0 0.0 0.0,
-1
-i
5 3 2 5 0 0 5 3 5 5 4
1 4 4
4 3 5 4
2 5 2 5
-
-
i
1.0 1.0 0.6 0.8 1.8 1.o 2.0 0.8 1.5 1.7 0.6 1.5 3.5 1.8 1.0 0.6
L
-
-
0.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 2.0 1.0 1.0 0.0
0.7 1.7 0.6 1.3 1.7
-
0.8 0.5 0.6 1.2 0.9 0.6 2.1 1.3 0.0 0.9
Maximum
-3
1.0
2.0
-3
'Difference. :Ricundylar width. Landmarks for measurement could not be located in 3-D CT reconstruction
TABLE 7. Variation between caliper and 3-D CT measurements
Measurement As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW2 Br-La Br-N U-D Enm-Enm FoRrd G-OP I-Gn La-O Mn-Mn N-A N-Gn N-Pcp NaRrd Ns-N Or-S
N
Zy-zy
SD
Minimum
A1
A
Li
(mm)
of h (mm)
(mm)
(mm)
1.8 2.p
1 .0 2.0
1.1
0.6
2.6 4.3
-
1.74 2.42
-
1.5 1.2 2.1 0.8 1.8
0.8 0.8 2.0 0.4 1.6
0.3 0.4 0.6 0.5 0.1
2.6 2.3 4.4 1.o 4.2,
1.43 3.35 5.49 0.58 1.62
-1
-1
-I
-1
-I
2.1 1.0 0.7 1.8 1.1 0.2 3.4 1.8
2.1 0.5 0.4 1.6 0.4
0.3 0.6 0.1 0.2 0.7 0.2 2.8 0.8 0.1 0.8 0.2 0.8 1.1 0.2 0.2 0.5
5.5 1.6 1.1 3.9 1.6 0.2 4.6 2.4 3.7 3.3 2.3 2.9 1.8 4.1 0.5 1.8
13.18 3.15 2.48 1.01 3.46 0.17 7.66 2.55 0.92 2.54 4.30 3.06 2.95 5.38 0.14 0.97
-
-
4
1.1
3
1.9 1.1 1.6 1.4
OrRrd
Zm-Zm
Mean
2 5
2.1 0.3 1.2
-
-
-
0.9 0.7 1.7 1.3 0.9 0.9 0.5 1.6 0.2 0.6
'1)ifference. 'Ricondylar width. "1,andmarks fur measurement could nut br located in 3-D C'I' reconstruction.
-
Maximum
-
Percent incongruence
-
292
C.F. HILDEBOLT ET AL
Fig. 3. Suture enhancement. The lower views are the same as those above except they have undergone suture enhancement. Note the enhanced visualization of the lambdoid and sagittal sutures.
Fig. 4. Gradient shaded 3-D CT reconstruction. Note improved ability to visualize landmarks compared with the image of Figure 2, a surface-shaded image.
293
3-D SKULL MEASUREMENTS
to locate landmarks in the X-Y directions, but it is hard to locate points in the Z direction. Dacryon (D) is an example of such a point. Because D is located on the medial orbital wall, it is hard to know whether one has actually placed the cursor on D or is anterior or posterior to D; that is, locating D in the X-Y direction can be done, but the 2 direction is problematic. The use of multiple images viewed independently only partially solves this problem in that one has t o remember the location of landmarks as depicted in the multiple views, and that is hard to do. We have found that most measurements can be made fromjust a few views and that additional views do not eliminate ambiguity but merely increase the time required to make measurements. The current study is limited to a consideration of measurements made between standard anthropological landmarks. Some of these are extremal landmarks (e.g., Gn is the lowest, V the highest, and Zy the most lateral) and as such are particularly difficult to locate on CT slices. By carefully selecting landmarks, it might be possible t o improve greatly the precision of measurements made from CT slices. Nevertheless, based on our pilot study, we believe that 3-D CT measurements are, in eneral, superior to measurements made rom the original CT scans. There is, however, considerable room for improvement in the precision of 3-D CT measurements. So far no one has determined how variations in imaging parameters affect 3-D CT measurement precision. Among the parameters that should be tested are variations in threshold levels, pixel size, radiation dose, head position, slice thickness, and table increment. Until these parameters are tested, it is impossible to know how to optimize measurement precision under current constraints or to know which parameters need further research attention. Methods also need to be ex lored for improving the appearance of 3-D T images. The ideal is for the 3-D CT image to simulate the appearance of a dry skull. Some of the recently developed gradient shading programs have significant potential to overcome feature visualization limitations (Geist and Vannier, 1988).Another development that would undoubtedly improve measurement precision would be to have multiple 3-D CT images displayed on the video monitor at the same time, along with multiple images of the cursor that is being used to locate landmarks.
f
8
In this preliminary study, we were concerned with the validity of basing commonly used anthropolo 'cal measurements on CT slices and 3-D T reconstructions. In the expanded studies, which are sug ested above, a more appropriate method o comparing methodologies would be to compare the coordinates of points (landmarks and fiducials) on objects with the coordinates of the corresponding points on images (the rationale for such an approach is described elsewhere; e.g., Cheverud and Richtsmeire, 1986). This comparison of coordinates involves the transformation of one coordinate system into the other. Such a comparison is beyond the scope of our preliminary study.
t?
!
CONCLUSIONS
If hi hly quantitative 3-D CT images can be pro uced and if the parameters responsible for the quantitative features of these images can be elucidated, an important tool would exist to im rove the current understanding of cranio acial dysmorphology. For example, a common component of craniofacia1 malformations is craniosynostosis. The pathophysiology of these disorders is poorly understood, largely because of the relative unavailability of pathologic material from deceased persons and because of inadequate imaging of defects in living persons. Improved anatomic fidelity of 3-D CT images would increase the database from which hypotheses regarding the pathophysiology of diverse craniofacial dysmorphology could be generated and tested. In this preliminary study, the repeatability of CT-slice and 3-D CT measurements were tested and compared with caliper measurements. In general, there were no statistically significant differences between the measurement techniques under investigation. There were, however, large discrepancies between caliper measurements and CTslice and 3-D CT measurements. Based on substantive consideration, we conclude that 3-D CT measurements are superior to measurements based on the original CT slices. There are, however, a number of parameters that need to be investigated before 3-D CT measurements can be considered optimal. There exists considerable otential for improving this modality for t e measurement of skulls. For the time being, this method should be used with caution in studies in which measurement errors of 5 mm or more cannot be tolerated.
%
P
R
294
C.F. HILDEBOLT ET AL ACKNOWLEDGMENTS
This study was supported in part by National Institute of Dental Research Grants DE08173 and DE08909. LITERATURE CITED Bass WM (1981) Human Osteology, 2nd ed. Columbia, MO: Missouri Archaeological Society. Brothwell, DR (1981) Digging Up Bones. Ithaca, NY: Cornell University. Broadbent BH Sr, Broadbent BH J r , and Golden WH (1975) Bolton Standards of Dentofacial Developmental Growth. St. Louis: CV Mosby. CheverudJM,and Richtsmeire J T (1986)Finite-element scaling applied to sexual dimorphism in rhesus macaque (Macaca mulattai facial growth. Syst. Zool. 35(3):109-127. Conroy GC, and Vannier MW (1984)Noninvasive threedimensional computer imaging of matrix-filled fossil skulls by high resolution computed tomography. Science 226:456-458. Conroy GC, Vannier MW (1987) Dental development of the taung skull from computerized tomography. Nature 329:625-627. Dye DM, andvannier MW (1984)Surface reconstruction from CT scans: Extended capability through spatial enumeration and bit Dlane encoding. In: Prof 8th Conf on Computers in Radiology, American College of Radiology, pp. 233-268. Gavou DE. Sammon NP. and Vannier MW (1984)Automatic generation of CAD models from CT data. In: Proc 8th Conf on Computers in Radiology, American College of Radiology, pp. 269-285. Geist D, and Vannier MW (1988) Medical imaging in a n object oriented environment. Proceeding of SPIE-the International Society for Optical Engineering, Vol914 (part 1of 2 arts), A plication of Optical Instrumentation in MeJicine Newport Beach California, (pp. 1299-1306). Hanson KM: Image rocessing (1985) Mathematics, engineering, or art? &:PIE Vol. 535 Application of Optical Instrumentation in Medicine 13:70-81. Hildebolt CF, and Vannier MW (1988)3-D Measurement accuracy of skull surface landmarks. Am J Phys. Anthropol. 76t497-503. Huijsmans DP, Lamers WH, Los JA, and Strackee J (1986) Toward computerized morphometric facilities: A review of 58 software packages for computer-aided three-dimensional reconstruction, quantification, and picture generation from parallel serial sections. Anat. Rec. 216r449-470. Knapp RH, and Vannier MW (1985)Generation of three dimensional images from CT scans: Technological perspective. Radiol. Technol. 56.391399.
d,
Marsh JL, and Vannier MW (1985) Comprehensive Care for Craniofacial Deformities. St. Louis: CV Mosby. Marsh JL, and Vannier MW (1987) The anatomy of the cranio-orbital deformities of craniosynostosis: Insights from 3-D images of CT scans. Clin. Plast. Surg. 14(1h49-60. Marsh JL, Vannier MW, Bresina S, et al. (1986)Applications of computer graphics in craniofacial surgery. Clin. Plast. Surg., 13(3):441-448. Marsh JL, Vannier MW, and Knapp RH (1985a) Computer assisted surface imaging for craniofacial deformities. In M Habal, et al. (eds):Advances in Plastic and Reconstructive Surgery. Chicago: Yearbook Medical Publishers, pp. 63-77. Marsh JL, Vannier MW, and Stevens WG (1984) Surface reconstructions form CT scans for evaluation of malignant skull destruction. Am. J. Surg. 148:530-533. Marsh JL,Vannier MW, Stevens WG, Warren JO, Gayou D, and Dye DN (1985b)Computerized imaging for soft tissue and osseous reconstruction in the head and neck. Plast. Surg. Clin. North Am. 12~279-291. McCormick BH, DeFanti TA, and Brown MD (1987) Visualization in Scientific Computing. Baltimore: ACM. Newton TH, and Potts DG (eds.) (1981) Technical Aspects of Computed Tomography. St Louis: C.V. Mosby. SAS Institute Inc. (1985) SAS User’s Guide: Statistics. Version 5 Edition. Gary, North Carolina: SAS Institute Inc. Stevens WG, Marsh JL, and Vannier MW (1984)Craniofacial surgical planning. Surg. Rounds March, pp. 2&27. Totty WG, and Vannier MW (1984) Analysis of complex musculoskeletal anatomy using three dimensional surface reconstruction. Radiology 132173-177. Vannier MW, Marsh JL, and Gad0 MH (1983a) Three dimensional display of intracranial soft tissue abnormalities. Am J Neuroradiol4:52&521. Vannier MW, Marsh JL, Stevens WG, et al. (1984a) Surface reconstruction and computer aided design of craniofacial surgical procedures based on CT scans. Proc Natl Comput. Graphics Assoc. 2:116-130. Vannier MW, Marsh JL, and Warren J O (1983b) Three dimensional computer graphics for craniofacial surGcal planning and evaluation. Comput. Graphics 17:263-273. Vannier MW, Marsh JL, and Warren J O (1984b) Three dimensional CT reconstruction images for craniofacial surgical planning and evaluation. Radiology 150: 179-184. Vannier MW, Totty WG, Stevens WG, Weeks PM, Dye DM, Daum WJ, Gilula LA, Murphy WA, and Knapp RH (1985) Musculoskeletal applications of three dimensional surface reconstructions, Orthop. Clin. North Am., 16.537-549, Weeks PM, Vannier MW, Stevens WG, and Knapp RH (1985) Three dimensional imaging of the wrist. J. Hand Surg. IOAr32-39.
Validation Study of Skull Three-Dimensional Computerized Tomography Measurements CHARLES F. HILDEBOLT, MICHAEL W. VANNIER, AND ROBERT H. KNAPP Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, Missouri 63110
KEY WORDS
Craniometry, Anthropometric, 3-D CT, CT
ABSTRACT
Recent advances in imaging have led to high-resolution computerized tomography (CT) scanning with exquisitely detailed slice images of the skull and three-dimensional (3-D) surface reconstructions using computer software. It is possible to use CT scans to acquire morphologic information about the skull in a convenient digital form and to derive 3-D measurements from surface reconstruction images. Unfortunately, no effort has been made to date to test the validity of these measurements on laboratory specimens, and no compelling evidence is available from phantom studies to indicate the nature and magnitude of the errors inherent in the measurement technique. We have performed a pilot study to quantify the morphology of the skull based on surface features that can be found in CT scans and 3-D reconstructions. Comparative measurements were obtained from five skulls (two normal and three with dysmorphology) with calipers and a 3-D electromagnetic digitizer. These measurements were statistically compared with those based on original CT scan slices and reformatted 3-D images. It is concluded that 3D-CT measurement techniques are superior to those in which measurements are obtained directly from the original CT slices; 3-D CT methods, however, must be significantly improved before measurements based on these techniques can be used in studies that require a high degree of precision. The results are used to indicate the most fruitful areas of future study.
In comparative anatomical studies, major concerns are 1) to quantify differences in forms and 2) to quantif form changes through time. Such quanti ication should be accurate, reproducible, and time efficient. Traditionally, quantification of skull form has been based on measurements made with simple rulers, calipers, and goniometers. The major limitations of these measurin devices are that they are time intensive an only surface measurements of external features are possible. Cephalometrics overcomes one of these limitations in that both internal and external structures (Broadbent et al., 1975) can be visualized; however, cephalometric radiographs are limited due to superimposition of structures, loss of skin surface detail due to X-ray transparency of soft tissues, and significant geometric magnification error due to the divergent nature of the X-ray beam.
P
3
@ 1990 WILEY-LISS, INC.
In the mid-l970s, computerized tomography (CT) scanning became available. Today several thousand CT scanners are in operation throughout the world. CT scans overcome man of the shortcomin s of conventional sku1 X-rays (Newton an Potts, 1981). They minimize superimposition of structures, possess high soft tissue contrast, and have essentially no geometric magnification error. A complete high-resolution CT scan examination of the head may, however, require more than 100 slices of 2 mm thickness each. It is difficult to work with such a large number of views and identify s ecific landmarks. Moreover, very few eop e are able to combine the slices menta ly into the 3-D images necessary to the evaluation of anatomical structure.
i
f
P f
Received June 20,1988; accepted October 16,1988.
284
C.F. HILDEBOLT ET AL.
To overcome these limitations, software programs were developed that transform CT scan slice data into a familiar and readily comprehensible 3-D format (Vannier et al., 1983a,b; 1984a,b; Marsh and Vannier, 1985). Methods and software were further developed to allow 1)the removal of overlyin anatomic structures, revealing deeper hi den structures and 2) the display of hard and soft tissue from any desired viewing perspective, including unobstructed views of the endo- and exocranial bases (Vannier et al., 1983a, 1984a,b, 1985; Dye and Vannier, 1984; Gayou et al., 1984; Marsh et al., 1984, 1985a,b, 1986; Stevens et al., 1984; Totty and Vannier, 1984; Knapp and Vannier, 1985; 1986; Weeks et al., 1985; Marsh and Vannier, 1987). Moreover, over 50 software packages have been develo ed for performing 3-D reconstructions o CT slices (Huijsmans et al., 1986). At many institutions, the use of 3-D CT reconstructions has become routine in qualitative evaluation of abnormalities of the head and in determinations of treatment protocols. There is growing speculation that 3-D CT may be particularly applicable to anthropometric and craniometric studies of prehistoric (Conroy and Vannier, 1984; Conroy and Vannier, 19871, living-, and deceased-contemporary populations, including Australasian populations. To date, however, no one has demonstrated the quantitative value of 3-D CT anthropometric measurements. This limitation is not readily ascertainable from the glowing descriptions often gwen in professional journals. It is clear that many people consider 3-D CT ima ‘ng to be a quantum advancement, but is t is the case? The ultimate goal in producing 3-D images is to convey information about objects to observers (Hanson, 1985; McCormick et al., 1987). The value of this information depends on its usefulness. A great deal of time, money, and thought have gone into using advanced technology to produce 3-D CT images. So far, most of the 3-D CT effort has been directed toward making the “pretty pictures” prettier. Terms such as “fidelity,” “ recision,” “quality,” and “artifact” have een used as artistic qualifiers rather than as scientific descriptors. Because the terms have not been defined with regard to 3-D CT and because statements involving these terms have not been supported with adequate quantitative data, evaluations of image-processing parameters
3
P
f
B
and judgments of image usefulness are of limited value. Moreover, without quantitation, one cannot com letely characterize norma1 and abnormal s ull morphology or measure the effects of growth or treatment. The study that we report is an initial step in determining the anatomical fidelity of 3-D CT images.
I:
MATERIALS AND METHODS
The objective of this study was to com are measurements based on original CT s ices and 3-D reconstructions with measurements made with calipers and validated with a 3-D digitizer system. For our study, we used the mean of original and repeat caliper measurements as truth and refer to the agreement between truth and CT-slice and 3-D CT measurements as “precision.” Standard spreading and sliding calipers were used to make the craniometric measurements, as given in commonly used osteological texts (Bass, 1981; Brothwell, 1981). The 3-D digitizer system (Hildebolt and Vannier, 1988) is an electromagnetic six-degree-of-freedom digitizer (3Space Digitizer/Tracker, McDonnell Douglas Electronics Company, Polhemus Navigation System Division, Colchester, VT) connected to a personal computer (Macintosh Plus, Apple Computer Corporation, Cupertino, CA) that was interfaced to a CAD/ CAM system (Unigraphics computer-aided design and computer-aided manufacturing system, McDonnell Douglas Manufacturing Systems Corporation, Cypress, CAI. Analyses were limited t o the measurements of bony anatomical landmarks of dry skulls. Landmarks and measurements are listed and defined in Table 1and are illustrated in Hildebolt and Vannier (1988).Additional information on landmarks is contained in Bass (1981) and Brothwell (1981). Five adult human skulls were selected for measurement. Two of the skulls were normal. The other three were from the museum collection of Dr. Paul Tessier (Paris, France) and represent the abnormalities of unicoronal synostosis, orbital neurofibromatosis, and nasofrontal meningoencephalocoele.The precision of the 3-D digitizer measurements compared well with caliper measurements, there being no significant difference (P < 0.05) between measuring techniques (Hildebolt and Vannier, 1988). CT scanning was performed with a standard clinical instrument with third-generation geometry (Somatom DR-H, Siemens
P
3-D SKULL MEASUREMENTS
285
TABLE 1. Landmarks and measurements used in study Landmarks A As B Br
D Enm Eu G Gn
I La Mn N Ns 0 Or OP PCP S V Zm ZY Measurements As-As BiW B-A B-Rr Br-N B-0 B-N B-Pcp B-V Br-La D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A NaBrd N-Gn N-Pcp Ns-N OrBrd Or-S zy-zy Zm-Zm
Alveolare, tip of maxillary alvolar bone between central incisors Asterion, junction of occipital, parietal, and temporal bones Basion, anterior midpoint of foramen magnum Bregma, junction of the sagittal and coronal sutures Dacryon, point at which frontal, maxillary, and lacrimal bones meet Endomolare, medial point of alveolar ridge a t the second molar The two Eu points establish a line that represents the greatest width of the skull Glabella, most anterior part of skull in midsagittal plane Gnathion, lowest point on mandible Infradentale, tip of mandibular alveolar bone between the incisors Lambda, junction of sagittal and lambdoid sutures; Mentale, most anterior point of mental foramen Nasion Nasospinale, highest point on nasal spine Opisthion, posterior midpoint of foramen magnum Orale, midpoint between central incisors on a line tangent to their posterior surfaces Opisthocranium, most posterior part of skull in midsagittal plane Posterior clinoid process Staphylion, point where a line tangent to the two curves a t the posterior of the palate crosses Vertex, highest point of skull Zygomaxillare, lowest point on the suture between the zygomatic and maxillary bones Zygion, most lateral point of zygomatic arch Biasterionic breadth Bicondylar width, distance between mandibular condyles. Basi-alveolar length Basi-bregmatic height Frontal chord Foraminal length Basinasal length Basiposterior-clinoid-process length Skull height Parietal chord dacryonic chord palatal breadth Foraminal breadth, maximum width of the foramen magnum. Skull length Symphyseal height Occipital chord foramen mentalia breadth Upper facial height Nasal breadth, maximum width of nasal opening perpendicular to sagittal plane Total facial height Nasio-posterior-clinoid-process length Nasal height Orbital breadth, maximum width of orbit measured from D Palatal length bizygomatic breadth maxillary breadth (midfacial breadth)
Medical Systems, Inc., Iselin, NJ). Two-millimeter, axial (perpendicular t o the long axis of the body), contiguous slices were made every 2 mm. These slices were parallel to the Frankfort horizontal plane, which passes through the most inferior margins of the bony orbits and the most superior margins of the auditory meatuses. Level-slicingcontour extraction was used to perform 3-D surface reconstructions. An observer (C.F.H.) recorded the X, Y, and Z coordinates of skull landmarks from CT slices, using the LI
(line), CO (column), and TP (table position), respectively (Fig. 1).The distances between landmarks were computed, with appropriate magnification correction factors being applied to the X and Y coordinates (when picture magnification was used). Measurements between landmarks were made directly from the 3-D CT reconstructions (Fig. 2) by means of an interactive computerized CT scan evaluation console (Siemens Evaluscope DR). Coordinates of landmarks, magnification factors, and 3-D CT distances
Fig. 1. Determination of coordinates of a point (see arrow on CT slice). The LI (line), CO (column), and TP (table position) represent, respectively, the X, Y, and Z coordinates of the point. TP represents the position ofthe skull in the scanner; that is, TP 1represents 1mm, and,
in this case, TP 79 represents the 79 mm from the starting position set by the operator on the CT scanner table. The LI and CO values are determined automatically with software that operates on the evaluation console.
Fig. 2. Illustration of dacryonic cord measurement on a 3-D CTreconstruction (see 1on image).The distance between landmarks (in this case, dacryons) is computed
automatically with software operating on the evaluation console.
3-D SKULL ME ASUREMENTS
287
were all determined with software that oper- several areas of the unicoronal synostosis, ated on a stand-alone CT scan evaluation orbital neurofibromatosis, and one of the console. All CT coordinate measurements normal skulls were not available for analysis were repeated over a 6 month period. from the 3-D CT reconstructions. Seven (5%) The percentage of variation between sets of the measurements, therefore, could not be of measurements (original vs. repeat) and made. Of the remaining 114 measurements between measurement techniques (caliper that could be made from 3-D CT reconstrucvs. CT slices and caliper vs. 3-D CT recon- tions, 27 (24%)could not be performed bestructions) were determined with the follow- cause landmarks could not be confidently ing formula. identified (Table 3). Results of CT-slice measurements and 3-D CT measurements are given in Tables 2 and 3. Standard deviations, N I minima, maxima, means of differences, and percent of incongruence between measurements (original vs. repeat, caliper vs. CT slices, and caliper vs. 3-D CT)were tabulated (Tables 4-7). The largest “mean difference” between where Xi, is a measurement in the firstset, ori ‘nal and repeat measurements made Xi, is a measurement in thesecond set, Xi, is wit TI CT slices was for FoBrd (2.2 mm; meanvalue for the first set, Xi, is mean value for the second set, and N is the number of Table 4). This is a 7.24% incongruence. Seven additional mean differences were measurements. The reliability of measurements (original 1mm or more. The largest “maximum differvs. repeat, caliper vs. CT slices, and caliper ence” between original and repeat CT-slice vs. 3-D CT) was assessed by calculating the measurements was 6.8 mm (FoBrd), with 11 means of the differences (between measures of the maximum differences exceeding2 mm. collected by the various methods) and the In the comparison of CT-slice and caliper standard deviations of the means of the dif- measurements, the largest mean difference 5.6 mm (Or-S; Table 51, which is a ferences (that is, we compared two sets of was 12.11% incongruence. Fourteen of the remeasures taken with the same instrument and then compared the mean of these mea- maining mean differences exceeded 2 mm. sures to similar means for other instru- The largest maximum difference was 9.2 mm ments). Student’s t tests and paired t tests (Zm-Zm), with 16 of the remaining maxidifferences exceeding 3 mm. (SAS Institute Inc., 1985) were used to test mum The largest mean differences between the null hypotheses that there were no diforiginal and repeat 3-D CT measurements ferences between the means of original vs. were for Or-S and As-As (3.5 mm; Table 6). repeat measurements and caliper vs. other These represented incongruences of 7.33% forms of measurement. The correlation coefficients for these measurements were also and 3.33%, respectively. Fourteen of the redetermined, as were the standard errors for maining mean differences equalled or exmodels with forced intercepts of zero and ceeded 1 mm. The largest maximum differslopes of one. For comparing techniques, the ence was 7.0 (As-As), with 17 of the averages of the two measurements (original maximum differences equalling or exceeding 2 mm. For variations between caliper and and repeat) were used. 3-D CT measurements, the largest mean difference was 3.4 mm (Mn-Mn;Table 7), for RESULTS a percent incongruence of 7.66. Four of the The top of the calvaria was missing from remaining mean differences exceeded 2 mm. the nasofrontal meningoencephalocoele The largest maximum difference was skull; so Br, V, and La could not be mea- 5.5 mm, with eight of the remaining maxisured. The calvariae of the unicoronal synos- mum differences exceeding 3 mm. tosis skull and one of the normal skulls were The results of Student’s t and paired t tests intact and Pcp could not be obtained with indicate that none of the differences between caliper measurements. Thus gold-standard the means of measurements were significant caliper determinations were not available (P > 0.05) except for Br-N for CT-slice vs. for nine (7%) of the measurements. Because caliper measurements. All the regressions of software and field of view limitations, were significant (P < 0.001) with correlaI
,
288
C.F. HILDEBOLT ET AL. TABLE 2. Original and repeat skull measurements f r o m CT slices
Measurement
Normal 1 (mm) 115 113 132 116 03: 132 121 124 109 026 034 037 204 033 099 048 070 121 030 054 051 041 089 136
As-As R-A B-Br B-N
B-0 B-Pcp Ba-V BiW' Br-La Br-N D-D Enm-EnLm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N Or-S OrBrd Zm-Zm Zv-zv
116 113 133 117 032 132 121 124 109 025 037 037 205 031 098 049 070 12J 030 055 051 039 089 136
Unicoronal synostosis (mm) 098 096 124 100 03z -
119 111 101 108 022 034 028 170 029 094 040 070
"9,024 052 042 037 098 131
Skull Meningoencephalocoele (mm) 094 093 08f! - 08; -
098 096 125 101
102 101 038 038 039 04t - -
03! 119 111 102 108 021 035 027 171 029 092 040 071 113
11:
-
024 052 045 038 097 131
11 : -
-L
-2
028 033 029 177 035 -
031 034 028 177 032 -
050 078 126 072 025 065 041 033 098 129
05 1 077 126 072 025 065 041 033 098 129
Orbital neurofibromatosis (mm) 105 092 128 095 036 036 123 111 125 110
016 031 029 176 022 084 043 059 099 075 025 046 040 033 078 127
105 096 127 096 034 035 121 108 125 110 015 031 026 178 023 086 043 060 099 074 025 044 039 044 078 126
Normal 2 (mm) 093 093 130 102 033 043 131 111 108 109 179 025 031 176 029 093 038 065 111 07 1 022 044 046 039 078 119
094 094 131 102 034 042 132 112 108 109 179 025 026 170 029 092 038 067 111 072 023 044 045 039 078 119
'Bicondylar width. > l o p of calvaria missing. "Calvaria intack; therefore, gold-standard mpasurrments not pvssihle.
2.
tions greater than 0.994 (the standard error for caliper vs. 3-D CT was 0.042 and for caliper vs. CT slices was 0.099). DISCUSSION
In general, we found no statistically significant differences between the measurement techniques under investigation; however, given the small sample sizes of our pilot study, type I1 errors may have been committed (that is, differences may actually exist). An expanded study is needed to clarify this point fully. At any rate, we suggest that there are substantive differences between the techniques, for the following reasons. Ideally measurements based on CT slices and 3-D CT reconstructions should be as precise as measurements made with calipers. In other words, the difference between caliper and CT-slice or 3-D CT measurements should not be any greater than are repeat Cali er measurements. In an earlier study, we etermined the amount of variation between original and repeat caliper measurements for the same 26 measure-
B
ments used in this study (Hildebolt and Vannier, 1988). In that study, the mean difference between original and repeat caliper measurements exceeded 1 mm for only one measurement. In our current study, when comparing caliper with CT-slice measurements, only four of the mean differences were less than 1mm, and the same was true for 3-D CT. Moreover, for CT-slice measurements, 15 of the mean measurements exceeded 2 mm, seven exceeded 3 mm, three exceeded 4 mm, and one exceeded 5 mm. 3-D CT measurements were considerably better, with only five measurements exceeding 2 mm and one measurement exceeding 3 mm. The superiority of 3-D CT to CT-slice measurements is also generally supported by the results presented above. We suggest, however, that there is considerable room for improvement with regard to the precision of measurements made from both CT slices and 3-D CT reconstructions. This is discussed below. There are problems with both techniques with regard to the time commitment re-
289
3-D SKULL MEASUREMENTS TABLE 3. Original and repeat skull measurements from 3 - 0 CT reconstructions
Measurement
Normal 1 (mmi
As-As R-A R-Hr R-N B-0 B-Pcp Ra-V BiW' Br-La Br-N
-2 111, 119 037 138 12if -
D-D
025 035 036 204 032 052 072 124,I 029 056 053 038 095 136
Enm-Enm FoBrd
G-Op I-Gn La-0 Mn-Mn N-A N-Gn
N-Pcp
NaBrd Ns-N Or-S OrHrd Zm-Zm Zy-zy
Unicoronal synostosis (mm)
-2 11; 116 03f 136 12f -
110 094 -5 099 036 036 -5 114
-J
-2 -2 027 038 037 201
04: 049 071 123 ,I 028 055 048 037 096 136
Skull Meningoencephalocoele (mm)
-5
027 028 169 169 03; 022 -
-
043 041 067 068 11: 11: 026 042 044 096 130
026 04,7 043 097 130
(mm)
-.I
-5
024 033 029 176 030
049 083 129 071 024 066 046 039
051 083 127 070 024 065 044 03J
-
Normal 2
107 093 -5 100 036 035 -5 114
025 033 029 177 029
-5
Orbital neurofibromatosis (mm)
-5
-
127 125
'Ricondylar width. 'Landmarks fnr measurement could not hc loreted in 3-D C T reconstruction. "Calvaria intack; therefore, gold-standard measurements not possible. :Missing due to field of view limitations. "Top of calvaria missing.
quired for optimal results. For CT-slice measurements, it is difficult to work with the large number of CT slices used in a scan of the skull (normal skull 1required 113 slices). It requires considerable time to look at each slice, and most people find it difficult to construct 3-D images of the skulls from memory. Such mental images are, however, necessary for one to be able to locate landmarks. Without exception, it is necessary to scroll back and forth among at least several slices to locate a landmark. In some instances (e.g., FoBrd, OrBrd, NaBrd, or EnmEnm), no amount of scrolling back and forth eliminates the ambiguity of identifying landmarks; that is, it is not possible to identify exactly which slice contains the desired landmark. With these limitations in mind, it is much easier to make measurements from 3-D CT reconstructions than from the original slices. Most landmarks, if visible on the reconstruction, can be located as quickly as they can on a dry skull, and the measurements can be
made even more quickly than with calipers. There are, however, two problems associated with locating landmarks on 3-D CT reconstructions. First, some points cannot be seen. For example, if the coronal and sagittal sutures are not patent, bregma cannot be seen. Suture enhancement, a prereconstruction processing procedure used to improve suture visualization, can be used to overcome this problem (Fig. 31, but this procedure requires that an operator edit all the CT slices that contain the involved sutures (Marsh et al., 1985a). This time commitment negates the time savings made by using 3-D CT reconstructions. It is also possible to enhance the sutures with gradient shading programs (Fig. 41, and in the future measurements will be made from these shaded images. The second problem with the 3-D CT reconstructions concerns locating landmarks on the computer display screen. If the X and Y axes describe the plane of the screen, and the Z axis is orthogonal (perpendicular) to the X and Y axes, then it is reasonably easy
290
C.F. HILDEBOLT ET AL TABLE 4. CT slice measurement variation Mean
Measurement
N
As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW' Br-La Br-N D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-PCD NaB;d Ns-N Or-S OrBrd Zm-Zm zv-zv
5 5 4 5 5 3 4 5 4 4
5 5 6 5 5 4 5 5 5 3 5 5 5 5 5 5
SD of A
Minimum
A' (mm)
(mm)
(mm)
0.8 1.0 0.7 0.9 1.0 0.9 0.6 0.8 0.5 0.3 1.2 1.2 2.2 1.9 0.6 1.3 0.4 0.9 0.2 0.7 0.7 0.4 1.o 0.6 0.2 0.3
0.8 1.7 0.3 0.6 1.2 0.3 0.9 1.0 0.5 0.2 1.1 1.1 2.8 2.4 0.9 I .0 0.4 0.9 1.3 0.8 0.8 0.7 1.2 0.5 0.4 0.4
0.0 0.0 0.4 0.0 0.0 0.5 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.2 0.0 0.1 0.0 0.0 0.0 0.0
A-A
0.1 0.0 0.0 0.0
Maximum (mm)
Percent incongruence
1.6 4.0 1.0 1.5 2.9 1.2 1.9 2.4 1.0 0.6 2.9 3.0 6.8 6.1 2.1 2.3 0.9 2.2 0.5 1.7 1.6 1.7 3.1 1.3 0.8 0.8
0.81 0.99 0.57 0.85 2.85 2.22 0.46 0.70 0.40 0.27 5.19 3.83 7.24 1.06 2.05 1.42 0.82 1.34 0.21 1.10 2.57 0.80 2.32 1.44 0.27 0.22
:Difference. Ricondylar width
TABLE 5. Variation between caliper and CT-slice measurements Mean
SD of A (mm)
Minimum A (mm)
Maximum
A'
(mm)
Percent incongruence
3.1 2.8 2.1 1.0 0.9 0.8 1.8 0.4 3.6 0.5 1.7 1.7 0.4 2.1 2.8 3.2 0.7 1.2 1.6 1.0 0.6 1.6 2.4 1.8 3.5 0.3
0.1 0.3 0.2 0.2 0.1 0.3 1.0 0.8 0.3 2.0 0.1 0.0 0.0 0.3 1.0 0.5 1.4 0.5 2.3 0.2 0.3 0.5 1.7 1.5 0.2 0.7
7.3 7.2 5.1 2.5 2.6 1.7 5.3 1.9 8.0 3.1 4.3 3.7 1.2 5.9 7.3 7.0 2.8 3.3 6.4 1.9 1.9 4.9 8.1 5.9 9.2 1.4
3.29 2.44 1.69 0.88 4.17 1.99 2.62 1.18 2.27 2.40 8.52 5.53 2.33 1.36 113 2 2.48 4.95 2.32 3.57 1.74 3.62 4.66 12.11 7.84 5.03 0.86
Measurement
N
(mm)
As-As B-A B-Br B-N B-0 B-Pcp Ba-V
5 5 4 5 5 ;3 5 5 4 4 5 5 5 5 5 4 5 5 5 3 5 5 5 5 5 5
3.4 2.4 2.2 0.9 1.5 0.8 3.3 1.3 2.6 2.6 1.8 1.8 0.7 2.4 3.4 2.3 2.1 1.6 4.2 1.3 0.9 2.4 5.6 3.0 4.5
RiW2 Br-La Br-N D-D Enm-Enm FoBrd G-Op I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N
Or-S OrBrd Zm-Zm Zy-zy :Difference. Bicondylar width
1.1
A
29 1
3-DSKULL MEASUREMENTS TABLE 6. 3 - 0 C T measurement variation
Mean
SD
Minimum
A'
A
A
(mm)
(mm)
Percent incongruence
-
7.0 1.o
3.33 l.0,2
-
-
3.0 2.0 1.0 2.0 2.0
1.38 2.26 1.77 76 89
2.0 3.0 1.0 3.0 4.0 1.0 3.0 1.o 2.0 3.0 2.0 2.0 5.0 4.0
4.93 2.96 2.02 0.44 5.36 1.18 4.40 1.04 1.26 2.32 2.41 2.83 7.33 4.56 1.04 0.47
Measurement
N
(mm)
of A (mm)
As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW Br-La Br-N D-D Enm-Enm FoBrd G-OP I-Gn La-0 Mn-Mn N-A N-Gn N-Pcp NaBrd Ns-N Or-S OrBrd Zm-Zm zv-Zv
2 4
3.5 1.o
5.0
0
0.0,
1.0,
0
-
5
1.4 0.8 0.7 1.0
0.9 0.8 0.6 1.4
1!.
0.7,
1.0 0.0 0.0 0.0 0.0,
-1
-i
5 3 2 5 0 0 5 3 5 5 4
1 4 4
4 3 5 4
2 5 2 5
-
-
i
1.0 1.0 0.6 0.8 1.8 1.o 2.0 0.8 1.5 1.7 0.6 1.5 3.5 1.8 1.0 0.6
L
-
-
0.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 2.0 1.0 1.0 0.0
0.7 1.7 0.6 1.3 1.7
-
0.8 0.5 0.6 1.2 0.9 0.6 2.1 1.3 0.0 0.9
Maximum
-3
1.0
2.0
-3
'Difference. :Ricundylar width. Landmarks for measurement could not be located in 3-D CT reconstruction
TABLE 7. Variation between caliper and 3-D CT measurements
Measurement As-As B-A B-Br B-N B-0 B-Pcp Ba-V BiW2 Br-La Br-N U-D Enm-Enm FoRrd G-OP I-Gn La-O Mn-Mn N-A N-Gn N-Pcp NaRrd Ns-N Or-S
N
Zy-zy
SD
Minimum
A1
A
Li
(mm)
of h (mm)
(mm)
(mm)
1.8 2.p
1 .0 2.0
1.1
0.6
2.6 4.3
-
1.74 2.42
-
1.5 1.2 2.1 0.8 1.8
0.8 0.8 2.0 0.4 1.6
0.3 0.4 0.6 0.5 0.1
2.6 2.3 4.4 1.o 4.2,
1.43 3.35 5.49 0.58 1.62
-1
-1
-I
-1
-I
2.1 1.0 0.7 1.8 1.1 0.2 3.4 1.8
2.1 0.5 0.4 1.6 0.4
0.3 0.6 0.1 0.2 0.7 0.2 2.8 0.8 0.1 0.8 0.2 0.8 1.1 0.2 0.2 0.5
5.5 1.6 1.1 3.9 1.6 0.2 4.6 2.4 3.7 3.3 2.3 2.9 1.8 4.1 0.5 1.8
13.18 3.15 2.48 1.01 3.46 0.17 7.66 2.55 0.92 2.54 4.30 3.06 2.95 5.38 0.14 0.97
-
-
4
1.1
3
1.9 1.1 1.6 1.4
OrRrd
Zm-Zm
Mean
2 5
2.1 0.3 1.2
-
-
-
0.9 0.7 1.7 1.3 0.9 0.9 0.5 1.6 0.2 0.6
'1)ifference. 'Ricondylar width. "1,andmarks fur measurement could nut br located in 3-D C'I' reconstruction.
-
Maximum
-
Percent incongruence
-
292
C.F. HILDEBOLT ET AL
Fig. 3. Suture enhancement. The lower views are the same as those above except they have undergone suture enhancement. Note the enhanced visualization of the lambdoid and sagittal sutures.
Fig. 4. Gradient shaded 3-D CT reconstruction. Note improved ability to visualize landmarks compared with the image of Figure 2, a surface-shaded image.
293
3-D SKULL MEASUREMENTS
to locate landmarks in the X-Y directions, but it is hard to locate points in the Z direction. Dacryon (D) is an example of such a point. Because D is located on the medial orbital wall, it is hard to know whether one has actually placed the cursor on D or is anterior or posterior to D; that is, locating D in the X-Y direction can be done, but the 2 direction is problematic. The use of multiple images viewed independently only partially solves this problem in that one has t o remember the location of landmarks as depicted in the multiple views, and that is hard to do. We have found that most measurements can be made fromjust a few views and that additional views do not eliminate ambiguity but merely increase the time required to make measurements. The current study is limited to a consideration of measurements made between standard anthropological landmarks. Some of these are extremal landmarks (e.g., Gn is the lowest, V the highest, and Zy the most lateral) and as such are particularly difficult to locate on CT slices. By carefully selecting landmarks, it might be possible t o improve greatly the precision of measurements made from CT slices. Nevertheless, based on our pilot study, we believe that 3-D CT measurements are, in eneral, superior to measurements made rom the original CT scans. There is, however, considerable room for improvement in the precision of 3-D CT measurements. So far no one has determined how variations in imaging parameters affect 3-D CT measurement precision. Among the parameters that should be tested are variations in threshold levels, pixel size, radiation dose, head position, slice thickness, and table increment. Until these parameters are tested, it is impossible to know how to optimize measurement precision under current constraints or to know which parameters need further research attention. Methods also need to be ex lored for improving the appearance of 3-D T images. The ideal is for the 3-D CT image to simulate the appearance of a dry skull. Some of the recently developed gradient shading programs have significant potential to overcome feature visualization limitations (Geist and Vannier, 1988).Another development that would undoubtedly improve measurement precision would be to have multiple 3-D CT images displayed on the video monitor at the same time, along with multiple images of the cursor that is being used to locate landmarks.
f
8
In this preliminary study, we were concerned with the validity of basing commonly used anthropolo 'cal measurements on CT slices and 3-D T reconstructions. In the expanded studies, which are sug ested above, a more appropriate method o comparing methodologies would be to compare the coordinates of points (landmarks and fiducials) on objects with the coordinates of the corresponding points on images (the rationale for such an approach is described elsewhere; e.g., Cheverud and Richtsmeire, 1986). This comparison of coordinates involves the transformation of one coordinate system into the other. Such a comparison is beyond the scope of our preliminary study.
t?
!
CONCLUSIONS
If hi hly quantitative 3-D CT images can be pro uced and if the parameters responsible for the quantitative features of these images can be elucidated, an important tool would exist to im rove the current understanding of cranio acial dysmorphology. For example, a common component of craniofacia1 malformations is craniosynostosis. The pathophysiology of these disorders is poorly understood, largely because of the relative unavailability of pathologic material from deceased persons and because of inadequate imaging of defects in living persons. Improved anatomic fidelity of 3-D CT images would increase the database from which hypotheses regarding the pathophysiology of diverse craniofacial dysmorphology could be generated and tested. In this preliminary study, the repeatability of CT-slice and 3-D CT measurements were tested and compared with caliper measurements. In general, there were no statistically significant differences between the measurement techniques under investigation. There were, however, large discrepancies between caliper measurements and CTslice and 3-D CT measurements. Based on substantive consideration, we conclude that 3-D CT measurements are superior to measurements based on the original CT slices. There are, however, a number of parameters that need to be investigated before 3-D CT measurements can be considered optimal. There exists considerable otential for improving this modality for t e measurement of skulls. For the time being, this method should be used with caution in studies in which measurement errors of 5 mm or more cannot be tolerated.
%
P
R
294
C.F. HILDEBOLT ET AL ACKNOWLEDGMENTS
This study was supported in part by National Institute of Dental Research Grants DE08173 and DE08909. LITERATURE CITED Bass WM (1981) Human Osteology, 2nd ed. Columbia, MO: Missouri Archaeological Society. Brothwell, DR (1981) Digging Up Bones. Ithaca, NY: Cornell University. Broadbent BH Sr, Broadbent BH J r , and Golden WH (1975) Bolton Standards of Dentofacial Developmental Growth. St. Louis: CV Mosby. CheverudJM,and Richtsmeire J T (1986)Finite-element scaling applied to sexual dimorphism in rhesus macaque (Macaca mulattai facial growth. Syst. Zool. 35(3):109-127. Conroy GC, and Vannier MW (1984)Noninvasive threedimensional computer imaging of matrix-filled fossil skulls by high resolution computed tomography. Science 226:456-458. Conroy GC, Vannier MW (1987) Dental development of the taung skull from computerized tomography. Nature 329:625-627. Dye DM, andvannier MW (1984)Surface reconstruction from CT scans: Extended capability through spatial enumeration and bit Dlane encoding. In: Prof 8th Conf on Computers in Radiology, American College of Radiology, pp. 233-268. Gavou DE. Sammon NP. and Vannier MW (1984)Automatic generation of CAD models from CT data. In: Proc 8th Conf on Computers in Radiology, American College of Radiology, pp. 269-285. Geist D, and Vannier MW (1988) Medical imaging in a n object oriented environment. Proceeding of SPIE-the International Society for Optical Engineering, Vol914 (part 1of 2 arts), A plication of Optical Instrumentation in MeJicine Newport Beach California, (pp. 1299-1306). Hanson KM: Image rocessing (1985) Mathematics, engineering, or art? &:PIE Vol. 535 Application of Optical Instrumentation in Medicine 13:70-81. Hildebolt CF, and Vannier MW (1988)3-D Measurement accuracy of skull surface landmarks. Am J Phys. Anthropol. 76t497-503. Huijsmans DP, Lamers WH, Los JA, and Strackee J (1986) Toward computerized morphometric facilities: A review of 58 software packages for computer-aided three-dimensional reconstruction, quantification, and picture generation from parallel serial sections. Anat. Rec. 216r449-470. Knapp RH, and Vannier MW (1985)Generation of three dimensional images from CT scans: Technological perspective. Radiol. Technol. 56.391399.
d,
Marsh JL, and Vannier MW (1985) Comprehensive Care for Craniofacial Deformities. St. Louis: CV Mosby. Marsh JL, and Vannier MW (1987) The anatomy of the cranio-orbital deformities of craniosynostosis: Insights from 3-D images of CT scans. Clin. Plast. Surg. 14(1h49-60. Marsh JL, Vannier MW, Bresina S, et al. (1986)Applications of computer graphics in craniofacial surgery. Clin. Plast. Surg., 13(3):441-448. Marsh JL, Vannier MW, and Knapp RH (1985a) Computer assisted surface imaging for craniofacial deformities. In M Habal, et al. (eds):Advances in Plastic and Reconstructive Surgery. Chicago: Yearbook Medical Publishers, pp. 63-77. Marsh JL, Vannier MW, and Stevens WG (1984) Surface reconstructions form CT scans for evaluation of malignant skull destruction. Am. J. Surg. 148:530-533. Marsh JL,Vannier MW, Stevens WG, Warren JO, Gayou D, and Dye DN (1985b)Computerized imaging for soft tissue and osseous reconstruction in the head and neck. Plast. Surg. Clin. North Am. 12~279-291. McCormick BH, DeFanti TA, and Brown MD (1987) Visualization in Scientific Computing. Baltimore: ACM. Newton TH, and Potts DG (eds.) (1981) Technical Aspects of Computed Tomography. St Louis: C.V. Mosby. SAS Institute Inc. (1985) SAS User’s Guide: Statistics. Version 5 Edition. Gary, North Carolina: SAS Institute Inc. Stevens WG, Marsh JL, and Vannier MW (1984)Craniofacial surgical planning. Surg. Rounds March, pp. 2&27. Totty WG, and Vannier MW (1984) Analysis of complex musculoskeletal anatomy using three dimensional surface reconstruction. Radiology 132173-177. Vannier MW, Marsh JL, and Gad0 MH (1983a) Three dimensional display of intracranial soft tissue abnormalities. Am J Neuroradiol4:52&521. Vannier MW, Marsh JL, Stevens WG, et al. (1984a) Surface reconstruction and computer aided design of craniofacial surgical procedures based on CT scans. Proc Natl Comput. Graphics Assoc. 2:116-130. Vannier MW, Marsh JL, and Warren J O (1983b) Three dimensional computer graphics for craniofacial surGcal planning and evaluation. Comput. Graphics 17:263-273. Vannier MW, Marsh JL, and Warren J O (1984b) Three dimensional CT reconstruction images for craniofacial surgical planning and evaluation. Radiology 150: 179-184. Vannier MW, Totty WG, Stevens WG, Weeks PM, Dye DM, Daum WJ, Gilula LA, Murphy WA, and Knapp RH (1985) Musculoskeletal applications of three dimensional surface reconstructions, Orthop. Clin. North Am., 16.537-549, Weeks PM, Vannier MW, Stevens WG, and Knapp RH (1985) Three dimensional imaging of the wrist. J. Hand Surg. IOAr32-39.
