Aesthetically ideal noses created using a single artificial intelligence model: Validating literature and exploring ethnic differences.

[BACKGROUND] Rhinoplasty remains one of the most popular cosmetic procedures worldwide and continues to evolve alongside shifting perceptions of beauty and identity. However, there are certain anthropometric standards that have emerged as the standard for preoperative assessment. As artificial intelligence (AI) continues to evolve, it can offer a unique perspective into understanding nasal beauty beyond these standards, especially in different ethnicities.

[OBJECTIVE] Our aim was to examine whether prompted aesthetically ideal AI-generated female noses reflect ethnic variation and align with published ideals, which may be relevant among preoperative discussions.

[METHODS] Using a publicly accessible AI diffusion model we generated 180 "aesthetically ideal" facial images for six ethnic groups: African American, Asian, Caucasian, Hispanic, Middle Eastern, and South Asian.

[RESULTS] The AI generated images demonstrated significant variation across the different ethnicities. Notably, only the African American cohort fell within the previously published alar base to tip ratio. Similarly, the Caucasian group aligned with the nose width to intercanthal distance ratio ideal. Interestingly, all ethnicities significantly differed from the previously published nasal tip projection ratio. In contrast, nasolabial angles across all groups remained within the accepted ideal range. Regarding nasofrontal angles, only the Middle Eastern and South Asian cohorts aligned with the published standard.

[CONCLUSION] AI has the capability of generating realistic female nasal images of varying ethnicities, incorporating morphological nuances. Our findings differ from certain aesthetic standards and support an individualized approach to rhinoplasty planning. AI-generate images may be utilized as a supplementary visual reference in discussions around patient identity. However, its role in intraoperative use remains unvalidated.