Page 109 - Read Online

Page 151 Shen et al. Art Int Surg. 2025;5:150-9 https://dx.doi.org/10.20517/ais.2024.71

INTRODUCTION
Plastic surgery is a creative field at its core that is driven forward by innovation in both the techniques and
devices used in aesthetic and reconstructive procedures. Breast reconstruction is a critical aspect of breast
cancer treatment with significant physical and psychological benefits for patients . Artificial intelligence
[1]
(AI), which encompasses various fields such as machine learning and natural language processing, can
drastically transform current breast reconstruction practices. Recently, many applications of AI, such as
imaging analysis and symptom monitoring, are gradually becoming tools for plastic surgeons, both inside
and outside of the operating room. As various AI technologies continue to emerge in the surgical realm, it is
important to assess how these budding applications can be implemented in breast reconstruction to further
improve patient outcomes. Here, we present a narrative review of the latest AI developments relating to the
preoperative, intraoperative, and postoperative phases of breast reconstruction.

METHODS
This narrative review synthesizes key findings from the current literature on AI applications in breast
reconstruction. A comprehensive literature search was conducted on PubMed and MEDLINE on August 1,
2024. Articles published on or prior to this date were included in the query. Keywords included MeSH
terms “artificial intelligence”, “mammaplasty”, “surgery, plastic”, and “augmented reality”. Given the
narrative nature of the review, no formal inclusion or exclusion criteria were applied. Articles were included
if they offered specific discussions and presented novel findings on the use of AI, including machine
learning (ML), Large Language Models (LLMs), and convolutional neural networks (CNNs) in breast
reconstruction. Priority was given to original research, clinical trials, and articles with the most up-to-date
information within the field. Articles that offered substantial insights or unique perspectives on
advancements, limitations, and clinical applications were emphasized. The included articles were
categorized into pre-, intra-, and postoperative applications. Recurring findings, including AR, AI-assisted
perforator mapping, symptom monitoring, and patient satisfaction and outcome initiatives, were grouped
and discussed.

AI IN PREOPERATIVE PLANNING
The gold standard for autologous breast reconstruction is the deep inferior epigastric artery perforator
(DIEP) flap. Computed tomography angiography (CTA) is essential for plastic surgeons to understand the
complex vascular architecture associated with the inferior epigastric artery and to select appropriate
perforators for autologous breast reconstruction with DIEP flaps. However, manual interpretation of CTA
by radiologists and plastic surgeons is labor-intensive and subject to variability.

As a promising alternative, AI has been deployed to automate CTA interpretation for DIEP flap planning. A
recent study published by Mavioso et al. tested the feasibility of using semi-automated software to detect
perforators for DIEP reconstruction. This AI application reduced the time spent on planning each case by
approximately two hours . While the software’s accuracy is lower for smaller vessels (< 1.5mm), it performs
[2]
equally well for larger vessels compared to manual selection. Another recently published study by Lim et al.
[3]
compared four LLMs in the interpretation of CTA scans . Certain models, like Bing AI (Microsoft,
Redmond, Washington, 2023), demonstrated superior accuracy and readability. However, as noted by the
authors, the reliability of the output by these models remained a significant challenge, as they can
sometimes provide irrelevant or “hallucinated” references. This limitation can detrimentally mislead
surgeons, indicating the need for further development and testing of AI in CTA analysis.