Exploring the Balance Between Artificial Intelligence and Human Expertise in Shaping Breast Reconstruction Outcomes: A Comparative Reflection Study

Abstract

<b>Background/Objectives</b>: Artificial intelligence (AI) has shown potential in patient education and integration into clinical decision support systems. However, its performance in counseling patients on breast reconstruction currently remains underexplored. This study's objective is to compare AI-generated answers with expert surgeon responses to common patient questions (derived from clinical scenarios) in domains like oncological justification, reconstructive options, and postoperative care. <b>Methods</b>: We realized an observer-blinded study using five real-world clinical scenarios in the field of oncologic and reconstructive surgery of the breast. Both ChatGPT-5 (October 2025 version) and a senior board-certified plastic surgeon responded to frequently asked questions, which were split into three domains: (1) oncological and surgical justification; (2) reconstruction options and outcomes, respectively; and (3) postoperative period. The answers were evaluated by another senior plastic surgeon using a four-grade ordinal scoring system (1 = unsatisfactory, 4 = excellent), which assessed accuracy, completeness, safety, nuance, and alignment with the current guidelines. <b>Results</b>: Across a total of 40 questions, the average AI response score was 3.1 ± 0.6. Domain-specific items scored lowest values for oncological justification (2.8 ± 0.7) and higher values for reconstruction options/outcomes and postoperative care (both 3.2 ± 0.4). No AI response was graded as unsatisfactory (score 1). Responses graded 4 (15%) were considered comprehensive, accurate, and patient-friendly. <b>Conclusions</b>: Globally, ChatGPT-5 provides satisfactory, readable, and medically accurate answers to basic patient questions on breast reconstruction, with a few limitations in nuanced oncological justification.

Ähnliche Arbeiten