British Journal of Radiology (2005) 78, 198-201
© 2005 British Institute of Radiology
doi: 10.1259/bjr/26977910
A comparative study of evaluation of radiographs, CT and 3D reformatted CT in facial trauma: what is the role of 3D?
A D Reuben, MRCS1,
S R Watt-Smith, MD2,
D Dobson1 and
S J Golding, FRCR1
1 Departments of Radiology, University of Oxford and 2 Maxillofacial Surgery, John Radcliffe Hospital, Oxford, UK
Correspondence: Dr S J Golding
 |
Abstract
|
|---|
A panel of 17 trainee maxillofacial surgeons viewed 23 cases of facial trauma, demonstrated on radiographs, axial CT images and three-dimensional (3D) reformatted CT images in a standardized viewing format under standard conditions. Their diagnostic interpretation, extraction of information relevant to surgical management and subjective evaluation of each modality were recorded by standard questionnaire and compared with a gold standard evaluation by a consultant radiologist and surgeon reading based on results of clinical management. There were clear and measurable differences in the viewers' evaluations of radiographs, CT and 3D reformatted images. Overall, surgeons showed more accurate diagnostic reading of radiographs and 3D reformatted images. This was in contrast to their subjective assessment of the clinical value of each modality, which showed a strong preference for 3D over all other techniques and for CT over radiographs. However the perceived benefit of axial CT images over radiographs was not reproduced on objective testing in this group; surgeons appear to perform less well in interpreting CT images than their subjective response to the modality would suggest. This work has supported the view that surgeons value 3D imaging as a front-line tool in the evaluation and management of selected cases of acute facial trauma. We have demonstrated that the perceived benefits of 3D reformatted CT to surgeons appear real.
|
Introduction
|
|---|
Facial injuries are clinically significant because they are often complex in nature and may have serious functional and cosmetic sequelae. This makes accurate diagnostic evaluation essential. Modern imaging modalities, especially CT, have been shown to be of value in the assessment and management of acute facial trauma [1–3].
Surgeons frequently need to make their own evaluation of the degree of skeletal disruption revealed by imaging studies when planning initial treatment of facial fractures. 3D images reformatted from CT offer a subjectively attractive medium for displaying skeletal lesions and the technique has attracted interest in the management of patients with facial trauma, among surgeons in particular [4–6]. However to what extent this corresponds to clinicians' objective performance in reading the images has not been studied in detail. This study was undertaken to investigate the objective performance of trainee maxillofacial surgeons in their diagnostic evaluation of facial fractures on different imaging techniques, including 3D reformatted CT, and to determine the degree of correlation with their subjective views of the clinical value of each.
|
Patients and methods
|
|---|
Preparation of images
Imaging records were selected from 16 patients with acute facial trauma undergoing treatment at the John Radcliffe Hospital, Oxford over a 2 year period. Cases were chosen from radiographs, axial CT studies and 3D images to reflect a representative spectrum of facial trauma, from simple localized fractures to complex transfacial injury. Overall, the study group comprised seven examples of radiographs, nine CT studies and seven cases of 3D reconstruction, treated for the purpose of reading as 23 separate cases (Table 1
).
Selected radiographs were occipitomental or occipitofrontal views, as appropriate (Figure 1). All CT studies were obtained using a Pace CT system (General Electric Medical Systems, Slough, UK), using contiguous 1 mm or 3 mm sections through the face, with a 20 cm field of view. Images were reconstructed on a high resolution algorithm using a 512 x 512 matrix, giving a pixel size of 0.39 mm. Representative sections numbering 10 to 15 in each case were selected from each patient to show the important areas of fracture (Figure 2). 3D images were reformatted from the CT data by surface rendering using a stand-alone processing system and software provided by the Medical Physics Department, University College, London (R Williams, www.medphys.ucl.ac.uk/mgi/workstat.htm). Thresholds were chosen manually in each case to show skeletal details optimally, and to minimize pseudoforamen artefacts (Figure 3).
View larger version (92K):
[in this window]
[in a new window]
|
Figure 3. Surface rendered 3D reformatted CT image showing a comminuted fracture of the left zygoma and maxilla.
|
|
Evaluation of images
A standard viewing format was created by transferring all images to high resolution VHS video tape. A commercial video camera (Sony Hi 8; Sony, Tokyo, Japan) was used to film radiographs on a light box. The selected CT sections were individually recorded by direct video link from the system computer to the video recorder, window width and level display having been set manually for optimal display of bony structures. 3D reformatted CT images were recorded directly from the video output of the processing unit and displayed in frontal view in rotating frame over 180°, in 3° increments, using a light source perspective technique. The video recording was edited on an editing station (Panasonic HD700; Matsushitu Electric Corp. of America, Secaucus, NJ) to create the final viewing format, in which radiographs were displayed for 30 s, each CT section for 10 s and each of the 3D images for 30 s.
A panel of 17 trainee maxillofacial surgeons was recruited from the Eastman Dental Hospital. All viewed the video tape in one session and their diagnostic evaluations and subjective assessments of the cases were collected by means of a standard questionnaire.
The questionnaire began with an introduction explaining what the viewing panel were asked to do. After this the panel viewed each of the 23 cases in turn. They were presented with a list of 19 structures in the face and asked to indicate whether or not they thought each structure was involved by fracture, and if so whether left or right sides, or both. After viewing all cases the panel completed the final section of the questionnaire, which sampled their subjective responses. They graded the ease of interpretation for each three modalities on a five point scale (5 easy, 1 difficult) and were then asked to indicate which of the three types of image they found easiest to view for (a) understanding the pathology and (b) making a diagnosis. Finally they were asked to make a binary decision whether each of the three techniques provided sufficient information for four purposes: diagnosis, assessing the extent of fracture, predicting complications and planning surgery.
Analysis of results
Viewers' responses to part 1 of the questionnaire (objective reading) were evaluated by comparison with a gold standard reading, made by a consultant radiologist and surgeon who had been involved in the clinical care of the patients. This reading was made on both the video presentation and the source images, in order to compensate for possible loss of information in the recording process. However the video reproduction was deemed independently by both expert viewers to represent the original data with acceptable accuracy. The objective reading of the viewing panel of each of the 19 structures potentially involved by fracture were scored against the gold standard reading, giving a mark out of 19 for correctness. The average scores for each imaging modality were calculated from these. To determine the effect of differences between viewers the marks from individual viewers were subject to two-way analysis of variance, and the marks for each modality were compared using Scheffes multiple comparisons. Finally the objective performance of viewers was compared with their subjective response to the questions in the second part of the questionnaire.
|
Results
|
|---|
Objective responses
Compared against the standard reading, viewers achieved best scores for reading 3D images, followed by their scores for radiographs, with reading of CT performing least well. Mean scores out of 19 from each candidate for 3D images (7 cases) ranged from 12.71 to 15.29 (mean 14.29, standard deviation (SD) 2.31). For radiographs (7 cases) the range was 12.14 to 15 (mean 13.62, SD 2.51) and for CT (9 cases) the range was 10.44 to 13.89 (mean 12.31, SD 3.86). Overall "correct" percentage scores were 75.7% for 3D images, 71.5% for radiographs and 64.7% for CT.
Two-way analysis of variance showed that there was no significant difference between the performances of viewers over the series of scores, indicating that interobserver error could be discounted. It showed additionally that there was significant difference between the scores for each modality (p<0.0001), subject to further analysis. This was achieved using Scheffes multiple comparisons, which showed that the differences in results between radiographs and CT and between CT and 3D reformatted CT were significant at the level of testing (p<0.001). The difference in results between radiography and 3D images was found to be insignificant (p<0.001), there being a 24% probability that this result could have happened by chance.
Overall, analysis of the objective responses indicates that surgical viewers perform better, in terms of recognition of known fracture lines, with both 3D reformatted CT and conventional radiographs, than they do when interpreting conventional CT images.
Subjective responses
All 17 viewers rated the ease of interpretation of each modality, giving an average rating of 5 (excellent) to 3D reformatted CT, 4 to CT and 3 to radiographs. Only 14 viewers indicated which modality they found to provide a better understanding of pathology; 13 (95%) felt that 3D images were best, and 1 thought CT. 12 of the 14 (86%) rated 3D images the best for clarity in making a diagnosis and 2 preferred CT. No viewer preferred radiographs for either purpose. The results showed a more positive subjective response to CT than had been suggested by viewers' objective performance, while reinforcing their overall preference for 3D reformatted CT. Viewers' responses when asked to indicate whether or not each technique provided enough information for diagnosis, assessing extent of injury, predicting potential complications and planning surgery are shown in Table 2.
View this table:
[in this window]
[in a new window]
|
Table 2. Viewers' opinions on the clinical role of each technique affirmative responses as a percentage of the total responses for each modality
|
|
The results found heavily in favour of 3D reformatted CT, with all 17 viewers reporting that 3D provided enough information on its own in categories 1, 2 and 4. CT was the next preferred modality in general but it was notable that only 50% of viewers felt that radiographs showed enough information for diagnosis, in contrast to their objective scores, and most found radiographs unsatisfactory for assessing fracture extent and planning treatment. All viewers rated each technique less well for predicting potential complications, compared with other clinical uses.
Overall the results show that subjectively viewers show a preference for 3D images over conventional CT and over radiographs, while the objective assessment shows better performance with 3D over conventional CT but also better performance on radiographs over conventional CT.
|
Discussion
|
|---|
This study demonstrates that surgical viewers find 3D reformatted images to be of the greatest clarity and easiest to interpret of the available imaging modalities in facial trauma. All of those questioned reported that the 3D images presented enough information for diagnosis of the fracture, for assessing the extent of the fracture and for planning surgical treatment in the patient. Statistical analysis confirmed that candidates' objective performance in interpreting 3D reformatted CT was significantly better than when interpreting conventional CT. The results compare favourably with previous studies showing a subjective preference for 3D images among clinicians [7–10] but in addition they documented a significant difference between subjective evaluation and objective performance in this group of surgeons. Objective testing showed that for this group 3D reformatted CT gave the most accurate performance, followed by radiographs and then conventional CT. Statistical analysis revealed the study to be robust and confirmed as significant the differences between 3D and CT and between radiographs and CT. On subjective testing surgeons reported 3D images to be the most easily interpreted, to best provide understanding of the pathology, and to be effective for predicting potential complications and for treatment planning.
A similar study by Alder et al revealed subjective preference for 3D over conventional modalities, which was reflected in an alteration in surgical approach in 45% of cases [7]. They concluded that 3D images are of greatest benefit for the assessment of mid-face injuries, although they recognized that disparity between those who reported the images to be useful (97%) compared with those who modified their diagnosis (28%) was large. Mayer et al [8] performed subjective and objective analysis of the diagnostic capabilities of 3D reformatted CT as compared with CT and radiographs, finding that 3D provided superior definition of fracture lines (especially horizontal ones) and better appreciation of the extent of comminution. Gillespie et al [1] in their study found 3D to be of greatest value in patients with severe trauma and multiple fractures but less useful in minor trauma where there was no fragment displacement.
A key feature of our results is the lack of correlation between viewers' subjective and objective results with respect to CT. Axial CT images may be held to contain more information than the other techniques but despite their subjective preference for CT over radiographs, surgeons actually performed less well in interpreting CT. These responses may reflect training and familiarity with different imaging modalities, giving a bias to radiographs in particular, which are more likely to be interpreted by surgeons themselves, whereas radiological support is more likely to be available in evaluating CT. If so, our results have implications to the training of junior maxillofacial surgeons, the group used in this study, who are most likely to be engaged in the first-line management of facial trauma. Clinical training and experience may explain the preference for 3D reformatted CT, as these images relate readily to structures with which surgeons are confronted during surgical procedures.
We wished to make an objective comparison of interpretation of different techniques and for this reason chose a standardized viewing method using the medium of videotape and television monitor. This is a familiar method of viewing 3D reformatted CT and may have introduced bias in favour of 3D, to the detriment of the other two techniques. Further study of our own work has also shown rotating frame display to enhance image interpretation. Additionally, the viewing format dictated selectivity of images, which may have introduced limitations in reading CT in particular, although a radiologist and surgeon experienced in this work were satisfied that selected images were a faithful reflection of the fracture in each case. An ideal evaluation would display images in their conventional formats and allow variable time for viewing and assimilation of information, as would be the case in a clinical scenario.
In order to compare techniques we chose to treat each radiograph, CT and 3D reformatted CT as a separate case. However we believe there is a case for a study of the effect of imaging in clinical management, in which all three modalities would be available for each patient and the contribution of each would be evaluated separately. The cases used in this study also varied greatly in their severity and anatomical distribution within the face. It has yet to be established which injuries are more appropriately evaluated with 3D reformation in particular and so for clinically relevant assessment cases could be grouped according to the severity of their injury.
Effective clinical imaging must always be justified by positive influence on clinical management. In facial trauma imaging provides information which is known to contribute to accurate diagnosis and greater understanding of the degree and extent of injury but ideally this should be reflected in reduced operating times, reduction in post-operative complications, shorter hospital stays and improved clinical result. Our study has reinforced the role of 3D reformatted CT in conveying essential information to clinicians dealing with facial injury but we believe that further work is needed to establish precisely where and when each imaging technique should be used for most clinical benefit.
|
Footnotes
|
|---|
Current address for Dr Adam D Reuben, Hillcrest, Netherclay, Bishop's Hull, Taunton TA1 5ED, UK. 
Received for publication January 5, 2004.
Revision received August 19, 2004.
Accepted for publication October 13, 2004.
|
References
|
|---|
- Gillespie JF, Isherwood I, Barker GR, Quayle AA. 3D reformations of CT in the assessment of facial trauma. Clin Radiol 1987;38:523–6.[CrossRef][Medline]
- Johnson DH. CT of maxillofacial trauma. Radiol Clin N Am 1984;112:146–53.
- Zilka A. CT in facial trauma. Radiol Clin N Am 1984;22:131–44.[Medline]
- Gillespie JE, Isherwood I. 3D anatomical images from CT scans. Br J Radiol 1993;59:289–92.
- Zonneveld FW, Lobregt S, van der Meulen JC, Vaandrager JM. 3D imaging in craniofacial surgery. World J Surg 1989;13:328–42.[CrossRef][Medline]
- Zonneveld FW. A decade of clinical 3D imaging, a review. Part 2: clinical applications. Invest Radiol 1994;29:574–89.[CrossRef][Medline]
- Alder ME, Deahl ST, Matteson SR. Clinical usefulness of two-dimensional reformatted and three-dimensionally rendered computerised tomographic images. J Oral Maxillofacial Surg 1995;53:375–86.[CrossRef][Medline]
- Mayer JS, Wainwright DJ, Yeakley JW, Lee KF. The role of three-dimensional computed tomography in the management of maxillofacial trauma. J Trauma 1988;28:1043–53.[Medline]
- Magid D, Michelson JD, Ney D, Fishman EK. Adult ankle fractures: comparison of plain films and interactive 2 and 3D CT scans. Am J Radiol 1990;154:1017–23.[Abstract/Free Full Text]
- DeMarino DP, Steiner E, Poster RB, Katsberg RW, Hengerer AS, Herman GT. 3D CT in maxillofacial trauma. Am J Radiol 1988;138:477–83.
This article has been cited by other articles:
|
|
|
|
 
BJR Review of the Year - 2005.
Br. J. Radiol.,
March 1, 2006;
79(939):
183 - 187.
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
