Clinical assessment of breast symmetry and aesthetic outcome: can 3D imaging be the gold standard?

Abstract There is a lack of an accurate standardised objective method to assess aesthetic outcome after breast surgery. In this methodological study, we investigated the intra- and inter-observer reproducibility of breast symmetry and volume assessed using three-dimensional surface imaging (3D-SI), evaluated the reproducibility depending on imaging posture, and proposed a new combined volume-shape-symmetry ( ) parameter. Images were acquired using the VECTRA XT 3D imaging system, and analysed by two observers using VECTRA Analysis Module. Breast symmetry was measured through the root mean square distance. All women had undergone bilateral risk-reducing mastectomy and immediate breast reconstruction. The reproducibility and correlations of breast symmetry and volume measurements were compared using Bland–Altman’s plots and tested with Spearman’s rank correlation coefficient. 3D surface images of 58 women were analysed (348 symmetry measurements, 696 volume measurements). The intra-observer reproducibility of breast symmetry measurements was substantial–excellent, the inter-observer reproducibility was substantial, and the inter-posture reproducibility was substantial. For measurements of breast volumes, the intra-observer reproducibility was excellent, the inter-observer reproducibility was moderate–substantial, and the inter-posture reproducibility was substantial–excellent. The intra-observer reproducibility of was excellent while the inter-observer reproducibility was substantial for both observers, independent of posture. There were no statistically strong correlations between breast symmetry and volume differences. The intra-observer reproducibility was found to be substantial–excellent for several 3D-SI measurements independent of imaging posture. However, the inter-observer reproducibility was lower than the intra-observer reproducibility, indicating that 3D-SI in its present form is not a great assessment for symmetry.


Introduction
Volume, shape, and symmetry are important factors that influence the aesthetic appearance of the breasts. Clinical assessment of these factors often requires different techniques, where results are heterogeneous with varying reproducibility [1][2][3][4][5][6]. As breasts are three-dimensional (3D) structures, conventional techniques such as panel assessment or photogrammetry using biostereometric evaluation, might miss important information as they are based on two-dimensional (2D) images.
Three-dimensional surface imaging (3D-SI) uses the concept of triangulation to obtain coordinates of the imaged surface based on stereophotogrammetry. With more than one camera angle, a software system can plot the coordinates of a surface image based on the intercepting points from different camera angles, and create a perception of depth which is lost in traditional 2D photography [7]. Breast volumetric measurements acquired through the Vectra XT 3D imaging system was recently described to have excellent reliability, with higher reproducibility than measurements obtained through magnetic resonance imaging (MRI) [8]. However, there is still a lack of a standardised objective clinical method to assess breast symmetry and aesthetic outcome after surgery. Previous studies investigating breast symmetry have used ratios of measurements between anatomical landmarks, or differences between sagittal 2D sections of the breasts obtained from 3D images [9][10][11]. With evolved and enhanced 3D imaging techniques, we are now able to make use of more information from the 3D surface images during analysis [12,13].
The aim of the study was to investigate if the VECTRA XT 3D imaging system could provide reproducible assessments of breast symmetry and aesthetic outcome. The 3D measurements obtained from 3D surface images of women standing in two different bodily postures were evaluated with respect to intra-and inter-observer reproducibility. A secondary aim was to investigate whether symmetry correlates with volume difference or not, and to propose and evaluate a combined volume-shape-symmetry (VSS) parameter as a potential future measure in the clinical assessment of aesthetic evaluation.

Study sample
During October 2016 and January 2017, a pilot study was performed on healthy volunteers to test and optimise a running schedule for the current study. Subsequently, 88 women with high hereditary risk for breast cancer who were part of a longterm psychosocial follow-up study after bilateral risk-reducing mastectomy (RRM) and immediate breast reconstruction (IBR) with submuscular permanent silicone implants published in 2019 were invited for 3D-SI [14]. All women were operated at Karolinska University Hospital in Stockholm between 1997 and 2010. For asymptomatic healthy women, bilateral RRM and IBR was performed. For women who had previously undergone breast conserving surgery or mastectomy due to breast cancer, a contralateral RRM and IBR was performed in addition to a complementary ipsilateral procedure when needed. Written informed consent was signed by all participants included in this study regarding publication of their data and 3D surface images.

Photography procedure
Women were imaged using the VECTRA XT 3D imaging system (Canfield Scientific, Parsippany, NJ) with a camera shutter speed of approximately 3.5 ms. The device was calibrated daily. Measures to preserve anonymity were taken. Prior to imaging, the photographer marked out four anatomical landmarks with a pen: the suprasternal notch, the left and right clavicle 7 cm from the suprasternal notch, and the xiphoid process ( Figure 1). 3D surface images were captured in two different postures: (i) arms abducted at 45 with palms towards the floor and (ii) hands placed on their hips, shoulders in resting position, palms resting on the iliac crest. The position of the women was adjusted according to the camera screen gridlines, and the camera height was adjusted if needed. Three images per posture were captured at the end of a normal exhalation, with a short break for resting and repositioning of the women between each image to resemble a new 3D-SI appointment, yielding six unique images per participant.

Procedure of measurements
The measurements were performed according to a protocol developed by one of the co-authors (FM). Two observers measured the 3D images independently. Observer 1 (LB) analysed the image from the 'first appointment' of all women before moving on to the image from the 'second appointment', simulating a scenario where image analysis was performed continuously after each appointment. Observer 2 (OL) analysed all six images of one woman before moving on to the next set of images of another woman, simulating a scenario where images of the same woman captured at different time periods were analysed consecutively in order to study postoperative changes. All assessments were combined and stored for statistical analysis.

Measurement of breast symmetry
Root mean square distance (d RMS ) is defined as the mean distance between corresponding coordinates of two image layers or two 3D image surfaces expressed as a length. This is so far the only way that 3D-SI can measure the breast symmetry. It uses the information from thousands of coordinates that describes the imaged breast surface generated from the intercepting points from several camera angels plotted in a xyz-coordinate system to create an electronic 3D surface image, which can be examined and rendered when imported to different software programmes. By utilising the mathematical properties of d RMS on a 3D surface image of a torso together with a copied and reflected 3D surface image of the same torso, the mean distance between the corresponding coordinates of the left and the right side of the torso could be calculated. d RMS would then act as a measure of how symmetrical the left side of the torso is in comparison with the right side, where a d RMS of 0 mm would implicate that the corresponding coordinates of the two 3D surfaces are overlapping, i.e., the torso has perfect symmetry.
Breast symmetry was assessed by importing the 3D surface images into the VECTRA Analysis Module software, where image adjustments were performed to position the breast surface images in line with a coordinate gridline prior to further analysis, i.e., the pre-marked vertical anatomical landmarks (the suprasternal notch and xiphoid process) were aligned where x¼ 0, while the horizontal landmarks (on the clavicles) were aligned where y¼ 0. The images were cropped to maintain a fully anonymous 3D-SI torso, without chin or arms visible. By using the tool 'paint area selection', the breast area of interest was set to encompass an area 2 cm below the suprasternal notch to the left and right breasts where the pectoral muscle meets the breast, along the anterior axillary line, to 2 cm below the inframammary fold ( Figure  2(a)). The breast area of interest was then copied and reflected in the plane where x¼ 0, overlapping identically at the coordinates where x¼ 0 on the original breast surface, before d RMS was calculated.

Measurement of volume
Returning to the original non-reflected breast surface layer in VECTRA Analysis Module, the right breast area of interest was erased with the vertical midline as the border, using the 'lasso tool'. 1 The volume was measured to an interpolated virtual chest wall surface (Figure 2(b)), expressed in cubic centimetres. The same was done for the left breast volume measurement but erasing the right breast area of interest. The volume measurements of each woman were performed by two independent observers once per breast for all six images per woman.
Proposal of a new combined volume-shape-symmetry measure d RMS has a physical meaning of the mean difference between the surfaces of the two breasts. The dimension of this quantity is a length. However, the absolute value of this length is not always easy to judge, i.e., whether this value is large or small. Thus, in order to judge the d RMS value, it is important to compare this quantity with another value of length of the breasts. Since the volume of the breasts is measured, it is possible to estimate a 'characteristic' length of breasts (d b ). As the shape of the breasts is complex, the length of the breasts can only be determined by simplifying the shape of the breasts to certain ideal shapes, for example, a half oval or a hemisphere. A half oval or any other shapes than a hemisphere involves two or more lengths to define; thus, we have suggested to use a hemisphere as the model shape of the breasts: where V is the mean volume of the left and right breast. Both d RMS and the characteristic diameter are lengths, and they have the same unit of a length (mm, cm or m, Figure 3). If not, the unit must be converted to the same one. The ratio between d RMS and the mean diameter (d b ) of the two hemispherical breasts is thus a non-dimensional quantity that represents the relative difference between the two quantities. A value of zero of this quantity means that the two breasts are identical.
The proposed parameter to assist in the interpretation of d RMS , VSS, is 1 minus the ratio between the d RMS and the mean diameters of the two hemispherical breasts. VSS ¼ 1 means that the two breasts are identical; VSS ¼ 0 means that the difference between the two breasts is 100% (the difference between the two breasts is nearly the size of the breasts).

Statistical analysis
The term reproducibility is used as an umbrella term for both the concepts of agreement (related to the measurement unit) and reliability (a dimensionless parameter ranging from 0 to 1) according to de Vet et al. [15] Descriptive statistics for each image and observer are presented as means, standard deviations, range (min-max) and variance components (VCs). Graphical results are presented as Bland-Altman's plots, where each observation is based on the mean of the measurements on each of the three measured 3D-SI per woman. The 95% limits of agreement were  . Illustration of the coordinates and data used to obtain the mean distance between two breast surfaces (d RMS ) and the characteristic diameter (d b ) of the left (blue) and the right (maroon) breast, calculated from the measured volumes of the breasts in VECTRA Analysis Module. d RMS is calculated based on measuring the shortest distance between more than 1000 points on one surface to the corresponding points on the other surface. calculated taking the repeated measurements into account as described below.
Agreement was estimated using the repeatability coefficient (RC) for measurements by the same observer. The RC was estimated as 1:96 Â ffiffi ffi 2 p Â ffiffiffiffiffiffiffiffiffiffi ffi r 2 error p , with the r 2 error estimated, using a one-way ANOVA model with a random patient effect [16]. Agreement between observers was estimated using 95% Bland-Altman's limits. To take the data structure into account, these limits were calculated as mean differences (bias) between , where the VCs were estimated using a mixed-effects two-way ANOVA model with random patient effect and fixed observer effect [17]. Intra-observer reproducibility (for each observer separately) was estimated as in the one-way models, and as r 2 patient þ r 2 observer r 2 patient þr 2 observer þ r 2 error in the mixed-effects two-way models. In the mixed effects two-way models, inter-observer reproducibility was estimated as The interpretation of levels of reproducibility for different intraclass coefficient (ICC) values was as follows: 0.00-0.20 poor, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial and 0.81-1.00 excellent to perfect reproducibility [8,18]. The estimates of agreement and reproducibility are presented together with 95% confidence intervals.
The correlation between breast symmetry and volume difference (DV ¼ VL À VR j j ), and VSS and volume ratio (V rat ¼ VLÀVR j j VLþVR ) were estimated using the non-parametric test Spearman rankorder correlation coefficient (q), with a statistical significance level of .05.
All statistical analyses were performed using Stata/IC 14.2 for MAC. In this software, the procedure mixed is used for the random effect models.
The study was approved by the Regional Ethics Committee in Stockholm (dnr 2015/735-31/4).

Results
Sixty-four (73%) women took part in the study. The images of six women were excluded due to corrupt files, thus 3D surface images of 58 women were analysed. Clinical data are presented in Table 1. The reproducibility of the raw data (d RMS and breast volumes) was analysed by prior analysis of the proposed combined parameter VSS. Both observers' results including RC, reproducibility, and VC for the raw data and VSS obtained from 3D surface images captured in two different postures are presented in Tables 2 and 3, respectively.

Inter-observer reproducibility
There was a small, but consistent systematic difference, when comparing the two observers' measurements (Table 2, Figure 4). Observer 1 measured overall smaller values compared to observer 2. The inter-observer reproducibility was substantial (ICC ¼ 0.60-0.63) for measurements of breast symmetry, and moderate-substantial (ICC ¼ 0.54-0.66) for breast volumes.

Inter-posture reproducibility
The inter-posture reproducibility was substantial (ICC ¼ 0.71) for breast symmetry, and excellent to perfect (ICC ¼ 0.78-0.88) for breast volume measured by observer 1, data not shown. The inter-posture reproducibility was substantial (ICC ¼ 0.72) for breast symmetry, and substantial-excellent to perfect (ICC ¼ 0.78-0.88) for breast volume measured by observer 2, data not shown. There were no systematic errors between the measurements obtained from the 3D surface images captured in the two different postures for either observer. No significant difference was observed between breast symmetry and breast volume measurements performed by the observers.

Correlation between breast symmetry and volume difference
There were no strong statistically significant correlations between breast symmetry and breast volume differences (Table 4).

Volume-shape-symmetry parameter
The intra-observer reproducibility of VSS was excellent to perfect for both observers (ICC ¼ 0.81-0.82), while the inter-observer reproducibility was substantial (ICC ¼ 0.61-0.62), independent of posture (Table 3). Based on measurements performed by observer 1, VSS values were slightly higher than VSS values by observer 2. The variability between the observers' measurements was smaller for higher VSS values than for smaller VSS values, i.e., the observers' measurements were less coherent for women with less symmetrical breasts. No statistically significant correlations were found between VSS and volume ratio (data not shown).

Discussion
Measurements acquired from the VECTRA XT 3D imaging system which describes breast aesthetic outcomes in terms of breast symmetry were found to have substantial-excellent to perfect intraobserver reproducibility, while the intra-observer reproducibility of breast volume measurements was excellent to perfect. The interobserver reproducibility was, however, moderate-substantial. The reproducibility of 3D-SI measurements did not seem to be affected by which posture the women were standing in during imaging. No strong statistically significant correlations were found between breast symmetry and breast volume differences, nor between VSS and volume ratio. VSS was found to have excellent intra-observer reproducibility and substantial inter-observer reproducibility after being combined from the raw 3D-SI measurements. 3D-SI of breasts was first described in the early 2000s and claimed to be a useful tool in postoperative asymmetry corrections, helping surgeons understand factors affecting breast shape and aesthetic outcome after mammaplasty [19,20]. This was questioned in later studies, and the technique has not gained broad acceptance amongst breast surgeons [6,10]. Most studies of reproducibility using 3D-SI techniques have investigated volume measurements, with varying results [21][22][23][24][25]. More recent studies have shown better usefulness of the 3D-SI system in breast surgery [12]. In agreement with our results, both studies showed better intra-observer reproducibility when measuring breast volume than breast symmetry.
One of the disadvantages with measuring breast symmetry using d RMS is the difficulties with positioning the women un-tilted and completely parallel with the cameras. To overcome or decrease the impact of this, four anatomical landmarks were pre-marked and used when adjusting the 3D surface image in the VECTRA Analysis Module prior to copying and reflecting the 3D surface images over each other. Nevertheless, there might still have been a component of human error since the adjustments of a 3D surface image in relation to what the software gridlines uses as the xyz-coordinates is carried out by an observer. Small differences in defined coordinates of point of reflection (the plane where x¼ 0) yield differences in superimposed breast surface images between different observers. This could explain the moderate inter-observer reproducibility. Standardisation of 3D acquisition in pre-and postoperative breast imaging has been suggested to improve imaging reproducibility [22]. The limits of agreement of the inter-observer reproducibility for breast symmetry were similar in size as previously described [12]. The standard error of measurement was larger for observer 1's measurements compared to observer 2's. One reason could be differences in the order of image analysis. Observer 2 might have had an advantage in repeating the adjustments of the torso in the xyzplane, in the selection of breast area of interest, and furthermore remembered the previous assessments of the same woman. However, both methods resulted in substantial-excellent to perfect intra-observer reproducibility for breast symmetry and breast volume measurements.
Previously, it has been described that images of women with arms at 90 were more reproducible compared to images with hands placed on their hips [22]. In our study, the inter-posture reproducibility was substantial-excellent to perfect, indicating that as long as patients are 3D imaged in the same posture, the choice of posture of having arms at 45 or with hands on their hips seems to be less important.  No statistically strong correlations were found between breast symmetry and breast volume difference. Previous studies have typically evaluated these two factors separately. The goal with the proposal of a new parameter VSS was to incorporate both measurements into an overall parameter evaluating aesthetic outcome of 3D structures, and to facilitate work and application of such a measurement in clinical practice. We did not find any statistically significant correlations between VSS and volume ratio, indicating that breast volume is independent of d RMS and VSS: This means that the volume measurements on its own do not give information about the breasts' shape symmetry, thus d RMS is of interest/ importance to take into consideration in the evaluation of breast aesthetic outcome using 3D-SI measurements. However, as described earlier, the value of d RMS is difficult to judge on its own, thus the introduction of a relative value (VSS) is motivated. This parameter had excellent to perfect intra-observer reproducibility, indicating that it might be a potential parameter in future research to illustrate an overall measure of the aesthetic outcome. However, further studies testing its clinical utility are needed. The reasons to why the inter-observer reproducibility was lower than the intra-observer reproducibility originates from the issues with using d RMS as a way of describing breast symmetry as discussed earlier.
A strength with this study was that we showed that aesthetic outcome in terms of breast symmetry and breast volume could be assessed with an objective method with moderate-excellent to perfect intra-and inter-observer reproducibility. Moreover, the reproducibility for both breast symmetry and breast volume measurements, i.e., the raw data, were investigated before introducing and analysing the reproducibility of a summated score (VSS). d RMS itself may not accurately reflect breast asymmetry. This combined parameter has the advantage of offsetting the effect of breast volume on d RMS and makes it usable across a wide range of breasts. Other strengths with this study were the relatively large cohort of study participants and the amount of 3D surface images acquired with a standardised imaging procedure (three images captured in two different postures for each woman, in total six images per woman). In addition, each breast was assessed individually with respect to 3D-SI breast volume measurements.
The study also has some limitations. Clinical assessment of the aesthetic evaluation in terms of breast symmetry and breast volume using 3D-SI is not a time-efficient method. Although it can be considered to be an objective method, there are certain subjective components involved [3,21,[25][26][27]. First, the skin marks on the women are manually placed, thus individual changes may influence the assessment results. Second, the breast area of interest is selected by the observer based on specified landmarks, but occasionally it may be difficult to define the inframammary fold or the boarder of where the pectoral muscle meets the breast depending on the woman's body constitution. Both these aspects influence the reproducibility of the method. A limitation with stereophotogrammetry itself is the inability to capture obscured objects, which for instance is a problem during image analysis of ptotic breasts. Detailed clinical data such as anatomical distances or implant volumes and shapes were not available during the data analysis in the current study. Nevertheless, implant shape has not been shown to be easily determined by surgeons in vivo [28]. The clinical importance of our results needs further evaluation. The size of clinically acceptable limits of agreement has not yet been investigated for 3D-SI measurements. Studies evaluating the magnitude of variabilities in breast symmetry and breast volumes estimated using 3D-SI that are clinically acceptable, compared to subjective ratings of aesthetic outcomes by the women themselves are also of interest to investigate. Generalisation of our results to other 3D-SI methods other than VECTRA XT 3D imaging system should be made with caution, as it is possible that the results are dependent on the specific imaging system.

Conclusions
The VECTRA XT 3D imaging system has potential to be applied in clinical practice as the intra-observer reproducibility of 3D-SI measurements related with aesthetic outcome was found to be substantial-excellent to perfect, independent on imaging posture in this methodological study. Breast symmetry did not show any statistically strong correlations with breast volume differences, strengthening the need for a parameter incorporating both measurements. The proposed new parameter (VSS) was not statistically significantly correlated with volume ratio of the breasts, which further stresses the need for a relative parameter to aid in the interpretation of d RMS and evaluation of the aesthetic outcome using 3D-SI measurements. VSS showed excellent to perfect intraobserver reproducibility, albeit further confirmation is needed prior any use of it as a measure of aesthetic outcome in clinical practice. The inter-observer reproducibility was lower than the intra-observer reproducibility throughout the study, indicating that unless we come up with an automated system, d RMS is not an ideal tool for the assessment of breast symmetry.