Preoperative Computer Simulation in Rhinoplasty Using Previous Postoperative Images

Bashiri-Bawil M, Rahavi-Ezabadi S, Sadeghi M, Zoroofi RA, Amali A.

What this paper says

A system that finds a past patient with a similar face from a database of 400 cases and transplants their actual postoperative nose onto a new patient's photograph achieved over 80 percent accuracy in a 20-patient pilot.

Overview

Simulation normally works by digitally sculpting a patient's own photograph, which reflects what the software allows rather than what the surgeon has achieved. This system instead searches previous cases for a similar face and uses that patient's real result to predict the new one.

Sections of note

Images were standardized, then facial landmarking used 68 points on the frontal view and 19 points on the profile view.
Facial features were calculated and image retrieval performed by measuring similarity between the new patient's feature vector and those in the database.
The nasal area from the retrieved postoperative image was swapped onto the corresponding region of the new patient's face.
The database used color profile photographs of 400 randomly selected patients, 360 women and 40 men, operated at a single tertiary hospital from 2010 to 2018.
Accuracy of the simulation exceeded 80 percent in a pilot study of 20 patients.
The authors describe the system as fast, easy to handle, reliable and accurate in both frontal and profile views.
They suggest uses in surgeon-patient communication and in training surgeons and residents.

What it means for a patient

The predicted result comes from an actual outcome the surgeon achieved, not from software sculpting.
Both the front and side views are simulated, which matters because the front view is the harder one.
Limits: 20 patients in the pilot, one hospital, and the abstract does not define how accuracy was measured.
Predictions are only as good as the database, so they reflect one center's results.

Why this paper matters

Simulations that show results a surgeon cannot deliver are a documented source of dissatisfaction. Grounding the prediction in real past outcomes addresses that directly. The pilot is small and the accuracy measure undefined.

Terms

Facial landmarking: Marking defined anatomical points on an image for measurement.
Feature vector: A list of numbers describing an image's measurable characteristics.
Image retrieval: Searching a database for the image most similar to a query.
Preoperative simulation: A computer-altered image showing a predicted surgical result.
Pilot study: A small preliminary study testing feasibility.

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

From the abstract

“Providing realistic and useful preoperative counseling based on a surgeon's knowledge and experience. Objective: Using previous preoperative and postoperative patients' images to predict the postoperative result of a new query patient. Design, Setting, and Participants: After preprocessing for image standardization,…”

Excerpt; the full abstract is on PubMed.