Artificial Intelligence in Breast Reconstruction: A Scoping Review of Pre-, Intra-, and Postoperative Applications

Artificial intelligence (AI) has become increasingly integrated into breast reconstruction, transforming preoperative planning, intraoperative guidance, and postoperative follow-up. AI tools have shown potential to improve patient counseling, standardize imaging analysis, and predict clinical outcomes. However, current applications need further clinical integration and validation.

A scoping review was conducted following Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines. PubMed, MEDLINE, and Embase were searched independently for studies published from 2020 onward, reflecting the surge in AI innovation. Search terms included breast reconstruction, machine learning, artificial intelligence, large language model, deep learning, and deep inferior epigastric perforator. Eligible original studies were categorized into pre-, intra-, and postoperative applications.

Of 496 records screened, 40 studies met inclusion criteria. Most addressed preoperative use (n = 28). Large language models such as ChatGPT consistently produced readable, accurate counseling content, improved shared decision-making, and supported informed consent generation. Imaging studies using AI-based 3-dimensional scanning or magnetic resonance imaging segmentation achieved high accuracy and reduced analysis time versus manual methods. Predictive models accurately predicted complications, donor-site morbidity, radiotherapy need, and patient dissatisfaction, enabling tailored risk mitigation. Intraoperative AI was used for real-time perfusion assessment through thermal imaging and as cognitive support tools. Postoperatively, large language models enhanced clarity of discharge instructions, whereas neural networks facilitated rapid symmetry evaluation.

AI is reshaping breast reconstruction by improving counseling, planning, and postoperative evaluation. Although evidence remains strongest in preoperative applications, intra- and postoperative use are rapidly emerging. Future efforts should prioritize multicenter prospective validation and workflow integration to ensure safe, reproducible clinical adoption.