Lack of solid scientific validation, subjectivity, and anecdotal nature of reporting functional, psychological, and esthetic outcomes are dominant characteristics in esthetic surgery literature. Contrary to other medical or surgical specialties, quality of care and success of cosmetic surgical procedures in terms of improved esthetic appearance are usually assessed with illdefined outcomes and subjective measures very often from the patients’ and/or providers’ perspective. Traditional methods for assessing esthetic improvements following surgery rely basically on patient’s satisfaction reporting and conventional rater groups. Naturally, patient’s satisfaction and self-esteem evaluation with various questionnaires is an important outcome measure. However, esthetic improvement both surgeons and patients have great difficulty in defining. On the other hand, evaluation by conventional rater groups suffers from high subjectivity and response and confirmation biases. There is an increasing demand for providing reliable data about any given procedure to optimize cosmetic outcomes in both esthetic and reconstructive surgery. Great efforts are being made nowadays for higher accuracy in measuring and evaluating results.
Biologically, beauty is something that the visual processing segment of the nervous system finds attractive. Attractiveness, on the other hand, is an essential attribute of beauty; although objectivity or subjectivity of attractiveness can be debated, recent studies support the notion that attractiveness is objective and quantifiable. It can be evaluated by esthetic judgment that relies on a set of cortical areas supporting high-level visual analysis. Interestingly, ganglion cells in the retina are arranged so as to increase the dark-light boundary, making the visual cortex particularly sensitive to contrast, which is perceived as an attractive stimulus. Contrary to what is widely believed, despite some ethnic and cross-cultural differences, a universal standard of attractiveness based on well-defined and measurable parameters and features does in fact exist.
Artificial intelligence machine learning has been proposed to objectively assess attractiveness. However, due to difficulty and poor efficiency of feature extraction methods of masculinity and femininity sexual dimorphism, averageness, symmetry, and skin color and texture, and inability to combine various features together, this technology, though promising, has been difficult to implement. Despite the vast majority of available artificial intelligence platforms, current application in plastic surgery is limited to assessment of only a very small number of “objective” end criteria, such as patient age or sex; development of algorithms to assess “subjective” measures such as attractiveness is far more complicated. A recent report describing the use of commercially available artificial software demonstrated a large gap between human and artificial ratings of attractiveness; this constitutes a critical limitation to the software’s efficacy for plastic surgery outcome assessments. Moreover, this approach is limited by the data available to collect the 3-dimensional point cloud as well as the expensive and extensive computing training required. Furthermore, given inaccessibility to software algorithms development processes and barriers in methodological understanding, together with other limitations, available artificial intelligence programs at present, although promising, may be best suited as an adjunct rather than alternative to conventional rater groups assessment.1
One might argue that beauty and attractiveness are intimately linked to a patient’s willingness to undergo cosmetic surgery that is difficult if not impossible to objectively quantify. Certainly, patient’s drive is highly influenced by appearance comparison that mediates the relationship between idealized bodies and subjective dissatisfaction with one’s body as well as several other factors such as social and family status, reduction of self-esteem by falling short of societal standards, and psychological insecurity. Subjectivity of relationship between attitudes and behaviors within human action as well as reasoned action and variance in behavioral intention of a given individual2 must not be confused with perception by others of attractiveness and beauty that can possibly be measured objectively. Despite apparently unsurmountable difficulties in defining and measuring beauty and attractiveness that are fundamentally 2 abstract concepts and a dynamic evolving interplay between time-constant biologic traits and continuous molding and adaptation, and despite the fact that plastic surgery is an art that transcends rigid scientific methodologies, by measuring human perception and not abstract attributes of attractiveness and beauty, it is possible to objectively assess outcome of any esthetic intervention and bring some science to the art of plastic surgery.
Measuring human perception by eye-tracking technology is a recent advancement that has just been introduced to plastic surgery. It has been widely used in marketing for the study of visual attention effect on in-store decision making. Since first described for reporting facial perception of infants with cleft lip and palate in 2009, its application in plastic surgery has gained in popularity. A recent review of the literature revealed 23 reported cross-sectional studies with the majority involving facial deformities and surgeries. It was most commonly utilized to assess individuals with cleft lip/palate, followed by facial deformities, prominent ears and breast reconstruction, and facelift, coronal synostosis and 1 orthognathic patient. Visual gaze was recorded with screen-based eye-tracking machines, head-mounted eye-tracker or eye-tracker glasses. Differences in observers’ attention to control versus condition were assessed in 19 studies and attention to preoperative versus postoperative images in 7 studies.3
Eye tracking has the potential to overcome many of the hurdles facing artificial intelligence. It allows objective analysis of gaze based on eye movements, and has the ability to assess observers’ visual preferences in an objective manner that can be used to quantify the emotional response to visual stimuli. It seems to hold promise for better understanding of beauty and attractiveness. The main constant of this technology is the human eye that exhibits an innate ability to fixate on the physical characteristics of objects or individuals. Considering the crucial role of viewers’ evaluation of visible changes in appearance following surgery, it has great potential in determining overall success of any plastic surgery intervention. It can accurately detect and objectively measure differences in viewers’ attention to different visual stimuli.
Observer’s gaze, a proxy for visual attention and fixation, is generally attracted faster, and visual fixation prevails for longer, when presented with unharmonious features. This has significant implications in social interactions, especially for those who are esthetically disfigured. Using eye-tracking patterns can provide a window into conscious and unconscious human preferences and can provide quantifiable and objective assessment independent of patient or provider bias. Thus, it is probably an excellent metric to assess outcomes of plastic and reconstructive surgery. Furthermore, this technology is a useful tool in preoperative patient counseling; it could potentially help patients to realistically assess their goals preoperatively. It has also been suggested that eye tracking could be useful for screening patients seeking esthetic surgery for body dysmorphic disorder as these patients may have distorted visual gaze patterns.
It must be noted that though the ideal approach with this technology should seek to simulate social interactions as closely as possible. Quality of data from eye-tracking studies is only as good as the quality of the images used. High-quality media with high-resolution pictures scaled to life-sized proportions are recommended. Images of only a portion or with concealed portions of the body part being evaluated create an unrealistic environment that distorts normal human attention and perception. It is thus important to have high-quality standardized and properly scaled images with unified nondistracting background or other elements such as presence of skin lesions or complex pigmentation patterns. Furthermore, videos at maximum resolution with a high frame-rate and high dynamic range with details more easily detected by observers may be better than still images at representing real-life social interactions. Nevertheless, it must be noted that regarding different light angles for 2-dimensional standardized photographic documentation there might be a blind range in light angle (0–30 degrees) that does not influence assessment. Increasing light angle past this threshold value to 60 degrees might have a statistically significant impact on visual perception.
To reduce any distraction of gaze and to ensure participant’s focus on the images depicted on the screen, testing environment must also be standardized and constant for all participants. In their study about breast symmetry, all analyses were performed in a noise-free environment in the same room and with constant daylight lighting conditions. Moreover, eye-tracking device was calibrated for each participant to ensure accuracy of measurements. Potential influence of factors such as socioeconomic status, medical history, or cultural background of participants on their visual perception may also be a complicating factor. Furthermore, because visual perception may vary among observers, and among young individuals, adults, and the elderly, large unified data must be obtained to reach valid conclusions.
At present, various eye-tracking devices in addition to different definitions for fixation time, automated optical inspection, and time given per image, greatly affect interpretation of results. With the same subjects, different definitions are likely to result in different findings. Moreover, observer factors, specifically observer positioning, fatigue, and cognitive biases, must not be overlooked; observers should be blinded to the goal of the study otherwise their attention may shift to the studied area. Furthermore, to best minimize biases, laypersons are best recruited; observers with more experience and plastic surgery background tend to focus on different areas of the face or the body. Priming bias and the way exposure to a first visual stimulus impacts a subsequent stimulus is also a factor worth considering. Ocular dominance and the fact that visual input from 1 eye may be preferred over the other must also be considered as it could influence the viewing pattern of some observers and may limit somehow the accuracy and specificity of eye-tracking analysis unless subjects’ ocular dominance is assessed before the investigation.
Although eye-tracking data cannot be directly translated into subjective perception of beauty and attractiveness as several studies imply that observers tend to focus visual attention on unique features, either attractive or nonattractive, technological advancement will continue to augment the standards used in eye-tracking research. Irrespective of several limitations at present, if used properly, eye-tracking technology has the potential to provide an objective assessment of esthetic outcome. It may help identify parameters that play a significant role in the perception of beauty. One can imagine a future that exploits artificial intelligence in association with eye-tracking technology in which an algorithm might eliminate the need for human observers. Construction of a tool capable of reliable, automatic assessment, and mimicking human gaze, based on a deep convolutional neural network, may not be a far-fetched science fiction fantasy.
Irrespective of this projected quantum leap into the future, attractiveness of a human figure is the result of complex interactions among many numerous factors. Even though esthetic results are fundamentally rooted in the perceptions of the patient, potentially objective measures may certainly offer reliable guidelines to plan and assess outcome of any procedure we may perform.1 Whether eye-tracking technology and/or artificial intelligence can replace 1 day the intuitive judgment by the human eye remains to be demonstrated. Although artificial intelligence is expected to exceed human performance in some complex tasks, we can claim with confidence and with the artistic sensibility and intuition of plastic surgeons that there will never be a perfect substitute for human consciousness and judgment.
1. Peck CJ, Patel VK, Parsaei Y, et al. Commercial artificial intelligence software as a tool for assessing facial attractiveness: a proof-of-concept study in an orthognathic surgery cohort. Aesthetic Plast Surg 2022;46:1013–1016
2. Hwang K. Use of the theory of reasoned action in plastic surgery. J Craniofac Surg 2021;32:e849–e850
3. Asaad M, Dey JK, Al-Mouakeh A, et al. Eye-tracking technology in plastic and reconstructive surgery: a systematic review. Aesthet Surg J 2020;40:1022–1034