Turn Your Vision into Reality-AI-Powered Pre-operative Outcome Simulation in Rhinoplasty Surgery

Knoedler S, Alfertshofer M, Simon S, Panayi AC, Saadoun R, Palackic A, Falkner F, Hundeshagen G, Kauke-Navarro M, Vollbach FH, Bigdeli AK, Knoedler L.

PubMed 38777929 Publisher (DOI)Free full text (PMC)

What this paper says

An artificial intelligence model trained on 3030 rhinoplasty patients produced simulated results that 101 observers could distinguish from real postoperative photographs only 52.5 percent of the time, barely better than chance.

Overview

Managing patient expectations before rhinoplasty depends on communication, and simulated previews are one tool for that. The authors trained a Generative Adversarial Network, a type of model where two networks compete to produce and detect synthetic images, on 3030 patients' before and after photographs. They then tested whether people could tell the generated images from the real ones.

Sections of note

Design: model development with a human discrimination test. Level of evidence III.
Training data: pre- and postoperative images from 3030 rhinoplasty patients.
101 study participants were each shown 30 preoperative patient photographs, then a pair consisting of the real postoperative image and the generated image, and asked to pick the generated one.
Participants: 48 men and 53 women, mean age 31.6 years.
Accuracy at identifying the generated image: 52.5 percent, standard deviation 14.3. Chance would be 50 percent.
Men identified the generated images more often than women, 55.4 versus 49.6 percent, p equal to 0.042.
Stated conclusion: the model predicted representations that were not perceived as different from real outcomes.

What it means for a patient

The simulated images were essentially indistinguishable from real results to ordinary observers.
A convincing simulation is not a prediction of what surgery will achieve on any individual nose. It reflects the average results in the training data.
Realistic previews raise a practical risk: a patient may treat the image as a commitment rather than an illustration.
Limits: observers were untrained rather than surgeons, and the test measured realism, not accuracy of prediction for a given patient.

Why this paper matters

Mismatched expectations are a leading cause of dissatisfaction after rhinoplasty, and existing simulation software is slow and operator dependent. An automated model producing believable previews could change consultations. Whether the previews match what surgery actually delivers is the untested question that matters most.

Terms

Generative Adversarial Network: a model where one network generates images and another judges them, improving together.
Simulation: a computer generated preview of a possible surgical result.
Preoperative consultation: the visit where surgery is discussed and planned.
Expectation management: aligning what a patient anticipates with what surgery can deliver.
Standard deviation: a measure of how spread out a set of values is.

Summary written by rhinoplasty.cc from the abstract, 2026-09-09; not medical advice. The authors' own abstract follows.

Abstract

Background: The increasing demand and changing trends in rhinoplasty surgery emphasize the need for effective doctor-patient communication, for which Artificial Intelligence (AI) could be a valuable tool in managing patient expectations during pre-operative consultations.

Objective: To develop an AI-based model to simulate realistic postoperative rhinoplasty outcomes.

Methods: We trained a Generative Adversarial Network (GAN) using 3,030 rhinoplasty patients' pre- and postoperative images. One-hundred-one study participants were presented with 30 pre-rhinoplasty patient photographs followed by an image set consisting of the real postoperative versus the GAN-generated image and asked to identify the GAN-generated image.

Results: The study sample (48 males, 53 females, mean age of 31.6 ± 9.0 years) correctly identified the GAN-generated images with an accuracy of 52.5 ± 14.3%. Male study participants were more likely to identify the AI-generated images compared with female study participants (55.4% versus 49.6%; p = 0.042).

Conclusion: We presented a GAN-based simulator for rhinoplasty outcomes which used pre-operative patient images to predict accurate representations that were not perceived as different from real postoperative outcomes.

Level Of Evidence Iii: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

Abstract as indexed by PubMed; the article is open access (PubMed Central).