Truace
Subscribe

Artificial Intelligence-Assisted Reconstruction Planning of Cheek Soft Tissue Defects: A Comparative Study Against Surgical Expert Consensus

Background: Reconstructive planning for cheek soft tissue defects requires expert assessment of defect size and location to preserve aesthetic harmony. Artificial intelligence (AI), particularly large language models (LLMs), has emerged as a potential decision-support tool in plastic and reconstructive surgery. This study evaluates the accuracy and reliability of ChatGPT-4o in proposing reconstructive approaches for post-excisional cheek defects. Methods: A comparative observational image-based study was perform…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2088· cite & verify
Artificial Intelligence-Assisted Reconstruction Planning of Cheek Soft Tissue Defects: A Comparative Study Against Surgical Expert Consensus
Three cases of plastic nasal surgery : for saddle-shaped nose, removal of entire nose, and arched Roman nose : a paper read before the Philadelphia County Medical Society, February 26, 1896 by Keen, William W. (William Williams), 1837-1932 Royal College of Surgeons of England. Public domain

In brief

Researchers gave ChatGPT-4o 46 standardized facial photographs of healthy volunteers on which cheek defects had been drawn to mimic tissue lost after tumor removal. Each image came with a standard prompt containing patient clinical data, and two experts rated the model's reconstruction proposals on a five-point scale across five domains.

The findings suggest large language models could assist surgeons in planning reconstructions, although the evidence comes from simulated defects rather than real patients. The authors state that expert supervision remains indispensable and that prospective clinical validation is needed before routine use.

Main points

  1. Average rating across five domains was 4.22 out of 5, with clinical accuracy highest at 4.88.
  2. Aesthetic result scored 3.96 and relevance of alternatives scored 3.65, the lowest of the five domains.
  3. Inter-rater agreement between the two experts was strong (ICC 0.75), with kappa 0.82 for safety.
  4. The model could not generate illustrative depictions of the flaps it proposed.

The gain

In a simulated image-based test, ChatGPT-4o's proposed cheek reconstruction plans earned high expert ratings for clinical accuracy and safety, suggesting a possible decision-support role for surgeons.

The rundown

The study used a comparative observational image-based design with 46 oblique facial photographs. Defects were manually delineated on healthy volunteers to simulate post-excisional tissue loss, and each annotated photograph was submitted with a standardized prompt.

Two experts independently scored responses on clinical accuracy, safety, aesthetic result, first reconstruction selection and relevance of alternatives. Agreement was measured with Cohen's kappa and intraclass correlation, including kappa 0.76 for first reconstruction selection.

The authors conclude the tool shows promise as an adjunct to decision-making in cheek reconstructive planning. They rate the work level III evidence and call for prospective clinical validation.

Sources

  1. Peer-reviewedAesthetic Plastic Surgery2026-10-20

The debate