Computer Imaging in Rhinoplasty Study

Rhinoplasty is one of the most common procedures performed by facial plastic surgeons. For many years, documentation of an individual’s nose and the relationships of the face has been carried out using various techniques of imaging from silhouettes, to photography, and now more recently digital imaging. Imaging is recognized as a standard tool in the practice of facial plastic surgery, and has major advantages to both the patient and the surgeon (1,2).
Advances in computer technology and software has permitted the use of digital modification of pre-operative imaging to help predict the outcome of a rhinoplasty (3,4). When used in a responsible, realistic fashion, the authors believe that computer imaging can be an educational experience for both the patient and surgeon. The computer images are not a guarantee of a result, but represent a common aesthetic goal to work towards. Despite the availability of this technology, many surgeons are reluctant to use it. They feel that it may give patients unrealistic expectations. The issue of medicolegal liability is often raised if the computer image results are not obtained (5,6).
The reporting surgeons in the study have been using computer imaging for 5 years. The primary objective of the study was to determine whether or not pre-operative digital image modification in rhinolplasty patients is useful in predicting post-operative outcomes. Secondary outcomes of interest included a comparison of the predictive values of frontal versus lateral views, and levels of agreement between patients and surgeons.

Methods

Patients undergoing rhinoplasty were collected in a prospective manner from 2000 to 2006 by 2 surgeons (PS and RR). Patients undergoing rhinoplasty surgery were offered inclusion in the study. Patients who entered the study consented to the use of photography for the study purposes. Pre-operative photos were taken on all patients to include frontal, lateral, basal and three-quarter views. Computer modification of the frontal and lateral images was performed using the Mirror Canfield Systems in consultation with the patient to help predict the outcome of the rhinoplasty. It was made clear that the images represented the goal of the surgery however there was no guarantee as to the match of the final outcome to the images. The computer imaging was always performed by the operating surgeon.
Post-operative images were taken and at this visit, patients were shown the pre-operative computer manipulated images in addition to their actual post-operative images. All patients had undergone only one operation by the surgeon. Patients were asked to rank the frontal and lateral images on a scale (1=poor match, 2=average match, 3=very close match, 4=identical). The surgeons performed the same scoring for the other surgeons’ patients. Informed consent was obtained pre-operatively for all patients including the potential use of their images for research, and their ratings of the images for research endeavors. Data was tabulated and statistical analysis of weighted kappa scores was performed using MedCalc statistical software (MedCalc Software, Broekstraat 52, 9030 Mariakerke Belgium. The Wilcoxon matched pairs test was used as a nonparametric test that compares two paired groups, with a predetermined p value of 0.05.

Results

From 2000 to 2006, 1540 patients were seen for primary rhinoplasty surgery. 112 patients undergoing rhinoplasty elected to participate in the study. Basic demographics of the patients can be seen in Table 1.
Post-operative imaging was performed between 6 months and 2.5 years after the rhinoplasty (mean 11 months). Patient and surgeon comparisons of frontal and lateral pre-operative computer manipulated images can be seen in figures 1 and 2 respectively. Frontal images were rated as poor, average, very close, and identical by patients 6.2%, 52.7%, 33.0%, and 8.0%, and by surgeons 2.7%, 61.6%, 34.8%, 0.9% respectively. Wilxcoxon matched pairs test revealed a two-tailed probability of p=0.23 for frontal view comparison. Lateral images were rated as poor, average, very close, and identical by patients 1.8%, 18.8%, 48.2%, 31.2%, and by surgeons 0.9%, 24.1%, 59.8%, and 15.2% respectively. Wicoxon matched pairs test revealed a two-tailed probability of p=0.0024 for lateral view comparison.
Overall patient’s frontal and lateral images were considered very close or identical 40.2%/76.8% of the time for patients and 34.8%/ 73.2% for surgeons respectively.
To test for levels of agreement, weighted kappa scores were calculated. The test of agreement shows a weighted kappa of 0.42 for frontal views of patients and surgeons, and 0.65 for lateral views. When testing for levels of agreement between patients frontal and lateral view, weighted kappa was 0.49, and for surgeons 0.89. Strength of agreement for kappa scores 0.41-0.60 is considered moderate, from 0.61-0.80 is considered good, and 0.81-1.00 very good.
Examples of pre-operative imaging, pre-operative computer manipulated imaging, and post-operative imaging are presented in figures 3 and 4.

Discussion

The study population’s demographics were similar to most rhinoplasty practices. The preponderance of females and a mean age of 28.3 years is in keeping with traditional practices.
Imaging was performed a mean of 11 months post-rhinoplasty. The authors recognize that some may perceive the 6 month images to be relatively early for assessing results, however most swelling and dynamic post-operative changes have minimized by 6 months.
Surgeons did not score their own patients in order to decrease the amount of bias associated with the results since better scores would likely reflect a better perception of success. In general, frontal image comparisons were less likely to result in identical matches than were lateral images, which is illustrated in figures 1 and 2. The frontal image is more difficult to assess because of poor depth perception in facial images, and more subtle relationships that are changed in the course of a rhinoplasty such as alar base and tip width, or columellar hang. Lateral imaging reveals more obvious and common rhinoplasty components that are manipulated during rhinoplasty, such as dorsal hump reduction and tip projection.
The Wilcoxon matched pairs test for frontal images comparing patient scores with surgeons scores revealed a two-tailed probability of p=0.2301, meaning that there is no reason to conclude that the overall median rankings differ. This is in contrast to the lateral images where p=0.0024. This significant difference can be explained by the concept that more useful, distinguishing information is visible on lateral images and thus differences in scores become more evident.
Overall, very close or identical scores were lower for frontal views by both patient and surgeon (40.2% and 34.8%), compared to lateral views (76.8% and 73.2%). These results are comparable to other rates described in the literature (3,7). In both cases, surgeons are more critical when comparing the images as revealed by lower percentages in both categories.
In terms of agreement between patient and surgeon, weighted kappa scores revealed a moderate strength of agreement for the frontal views (weighted kappa 0.422), and good strength of agreement for the lateral views (weighted kappa 0.650). This illustrates that there is a better level of agreement on lateral views which is important in consideration of the fact that the surgeon’s judgment should coincide with the patient’s judgment.
There are limitations of the study which are to be recognized. Post-operative images were taken as early as 6 months from surgery which may be controversial in terms of revealing final outcomes; however this would have the effect of underestimating agreement. Since there were two surgeons involved in the study, it is possible that overall scoring definitions of ‘very close’ or ‘poor’ may be different, and could be reflected in outcome scores. This information is useful to those who currently pursue the practice of simulation images, however it is not applicable to those surgeons who do not use this type of computer software. Ewart et al have suggested good results with the use of less expensive computer software than the traditional software currently being used by the authors and other facial plastic surgeons, which may encourage more surgeons to adopt this practice for the benefit of them and their patients (8).

Conclusion

Computer image manipulation in rhinoplasty has several advantages. Our study shows that ratings by patients and surgeons comparing pre-operative manipulated images and post-operative outcomes is predictive in approximately ¾ patients, and that lateral images are more useful for this goal. We also show that the level of agreement between patient and surgeon scoring is good for lateral images and moderate for frontal. The use of digital image manipulation is an excellent tool, where advantages outweigh the disadvantages, and can have a positive impact on a rhinoplasty practice.

References