A recent research project featured in the Aesthetic Surgery Journal Open Forum reveals that an artificial intelligence (AI) tool can generate lifelike images of what many individuals perceive to be the ideal female breasts, noting that these perceptions can differ significantly among various racial groups. The study utilized an online image generator to create pictures of women from Caucasian, African American, and Asian backgrounds with “perfect” breasts, uncovering distinct variations in breast shape, nipple orientation, and other defining features. These findings indicate that beauty standards are not uniform and could be valuable for plastic surgeons in aligning their work with the racial and cultural identities of their patients.

The ongoing debate over what defines an “ideal breast” has been an enduring topic in plastic surgery and related fields. While many studies have attempted to identify specific characteristics and proportions, the essence of beauty remains subjective, shaped by a myriad of influences, such as race, culture, and personal preferences. The researchers noted that a substantial amount of existing literature has predominantly narrowed its focus to the ideals upheld by Caucasian women, which may overlook the complexities of beauty standards among different racial demographics.

Curious about the role of AI in reflecting diverse perceptions of ideal breast shapes, the researchers aimed to determine whether an AI-powered image generator – trained on extensive datasets – could produce images aligned with established aesthetic guidelines and if it could similarly generate varying breast forms for women from different racial backgrounds.

“As an aesthetic surgeon, I am deeply intrigued by how humans interpret aesthetic ideals of the body, whether it pertains to the face, breasts, or overall physique. This pursuit of aesthetic enhancement drives our surgical goals,” said Aaron Lee Wiegmann, the study’s author and chief resident in plastic and reconstructive surgery at Rush University Medical Center (@doctor.wiggy).

“Historically, literature on plastic surgery has largely concentrated on aesthetic ideals within the Caucasian community, but I believe there are subtle variations among different racial groups that need recognition. In aesthetic surgery, it’s essential for outcomes to resonate with the patient’s racial identity, as the concept of an ‘ideal breast’ may differ significantly across racial lines.”

The investigation employed a publicly accessible AI image-generation platform that enabled the input of text prompts to produce corresponding images. The researchers procured a commercial license for publication purposes, crafting tailored text prompts to guide the AI in generating images of women with aesthetically pleasing breasts.

The foundational prompt read: “A topless Caucasian woman with perfect aesthetically ideal breasts standing in three-quarter profile view.” Slight modifications to this prompt allowed for the generation of images depicting women of various races and perspectives. For images of African American and Asian women, the term “Caucasian” was substituted appropriately, and the view was altered from “three-quarter profile view” to “frontal view” to capture different angles.

The AI models utilized by the platform leverage advanced technology that learns to create images based on extensive image-text datasets. These models produce realistic representations of individuals based on descriptive text, designed for general public usage rather than medical applications. This factor was pivotal as it indicated that the AI-generated outputs would likely reflect broader societal beauty standards rather than specific medical or aesthetic guidelines.

After generating a diverse set of images, the researchers selected high-quality images fitting specific criteria for analysis. These criteria emphasized clarity in the image, visibility of breast borders and the nipple-areola complex, and avoidance of extreme angles or any visible distortions. Ultimately, they selected twenty-five images for each racial group (Caucasian, African American, Asian) in both profile and frontal views, totaling one hundred fifty images for their thorough evaluation.

Using image-editing software, the team conducted detailed measurements on the chosen images. To maintain consistency, measurements were represented in pixels, ensuring a standardized unit of measurement across all images. Following established protocols in breast aesthetic studies, three-quarter profile images were analyzed for breast shape.

Measurements assessed included the upper-to-lower pole ratio (the proportion of breast above and below the nipple), nipple angle (the direction the nipple points relative to the body), and subjective evaluations of the upper pole slope and lower pole convexity. Frontal images were employed to analyze the positioning of the nipple-areola complex and proportions of the nipple and areola, considering factors such as the nipple-areola complex position relative to breast borders and ratios of areola to breast width and nipple diameter to areola diameter.

Overall, the AI-generated images of breasts appeared realistic and cosmetically appealing across all racial groups, showcasing good size, projection, and a natural, gently drooping shape. The researchers noted, however, that a few images displayed breasts that might seem disproportionately large for the woman’s body frame, and occasionally depicted improbably muscular physiques.

Despite the overall similarities, statistical analyses indicated notable variations in specific breast traits among racial groups. Caucasian breasts exhibited tendencies towards a smaller upper structure and a larger lower portion, with nipples inclined upwards and a higher occurrence of a concave upper slope.

Conversely, the African American and Asian breast images reflected a larger upper segment combined with a smaller lower segment, with nipples pointed more frontally and a greater prevalence of a convex or straight upper slope. Additionally, the nipple-areola complex positioning diverged, with Caucasian breasts noticeably lower on the torso than those of their African American and Asian counterparts, and African American breasts demonstrated relatively larger areolas proportionate to breast width compared to Caucasian breasts.

“I was amazed by the consistent and quantifiable differences in breast appearance produced by AI for various racial demographics,” Wiegmann expressed to PsyPost. This study underscores the reality of aesthetic variations perceived by different racial groups, suggesting that plastic surgeons should engage patients in dialogue regarding these nuances during breast enhancement consultations.

In comparing the AI-generated breasts with previously recognized aesthetic ideals, the researchers discovered that the Caucasian breast representations closely matched traits typically labeled as ideal within existing plastic surgery literature, including similarities in upper-to-lower pole ratio and nipple angle.

However, the African American and Asian breasts diverged from these Caucasian-centric ideals, further emphasizing the observed racial diversity. Interestingly, existing literature proposes an ideal breast shape for Asian women characterized by a considerably larger upper section, but the AI-generated images for Asian women, while displaying a larger upper portion than those of Caucasian women, did not conform entirely to this traditionally large upper standard.

“The general public should be aware that AI possesses the capability to create photorealistic images of aesthetically appealing breasts that reveal significant racial distinctions,” Wiegmann stated. “Understanding this, AI holds tremendous potential for aiding patients in clarifying their objectives for breast enhancement surgeries. Additionally, AI may transform pre-operative simulations of prospective post-surgical results for patients.”

However, this potential also raises ethical concerns regarding its application in plastic surgery, including the risk of presenting misleading before-and-after images or fostering unrealistic beauty expectations.

“Patients must be cautious as unscrupulous surgeons might utilize AI-generated images for self-promotion, which could result in misleading representations of potential results, thereby inflating patient expectations and raising critical ethical concerns in the field of plastic surgery,” Wiegmann cautioned.

The researchers acknowledged some limitations in their study. Although measurements were uniformly applied, a degree of subjectivity was involved in pinpointing precise breast landmarks within the images. They also faced challenges controlling factors like lighting and positioning in the AI-produced images, which might have subtly affected the measurements.

A significant limitation is the ambiguity surrounding the datasets used for training the AI models. The researchers lacked clarity regarding the specific images and potential biases in the datasets, which could influence the resulting outputs. Nevertheless, the congruence of AI-generated Caucasian and Asian breasts with some previously acknowledged racial breast ideals lends reassurance about the findings’ applicability.

Future research could extend this study by examining AI’s perceptions of ideal breast shapes concerning various body types within each racial group. Investigating preferences for breast aesthetics in women of varying body compositions—thin, overweight, and obese—could reveal additional nuanced differences. Moreover, future analyses could delve into potential biases ingrained in AI training datasets for deeper understanding of how these biases may influence beauty standards.

“Our long-term objective is to harness AI, alongside its continuous advancements, to gain a better grasp of what diverse human groups perceive as the ideal aesthetic form, ultimately aiding plastic surgeons in achieving excellent surgical outcomes that reflect the preferences of both the patients and the practitioners,” Wiegmann concluded.

The study, titled “Aesthetically Ideal Breasts Created With Artificial Intelligence: Validating the Literature, Racial Differences, and Deep Fakes,” was co-authored by Aaron L. Wiegmann, Elizabeth S. O’Neill, Sammy Sinno, and Karol A. Gutowski.