Future of artificial intelligence in plastic surgery: Toward the development of specialty-specific large language models

Abstract

The recent advancements in artificial intelligence (AI) and large language models (LLMs) have generated significant interest within the plastic surgery community.1Farid Y. Fernando Botero Gutierrez L. Ortiz S. et al.Artificial Intelligence in Plastic Surgery: Insights from Plastic Surgeons, Education Integration, ChatGPT's Survey Predictions, and the Path Forward.Plast Reconstr Surg Glob Open. 2024; 12e5515https://doi.org/10.1097/GOX.0000000000005515Crossref PubMed Scopus (1) Google Scholar While current commercially available LLMs like ChatGPT (OpenAI, CA, USA), Claude (Anthropic, CA, USA) and Gemini (Google, CA, USA) have demonstrated impressive language generation capabilities, their lack of domain-specific training raises concerns about their reliability and applicability in clinical settings.2Parisi G.I. Kemker R. Part J.L. Kanan C. Wermter S. Continual lifelong learning with neural networks: A review.Neural Netw Off J Int Neural Netw Soc. 2019; 113: 54-71https://doi.org/10.1016/j.neunet.2019.01.012Crossref PubMed Scopus (1548) Google Scholar To truly leverage the power of these technologies in our surgical specialty, it is imperative that we develop our own LLMs specifically trained on peer-reviewed plastic surgery literature, clinical guidelines, and clinically validated datasets, fine-tuned with expert input to meet our unique clinical needs. One of the main limitations of current commercially available LLMs is the uncertainty surrounding the data on which they are trained. These models are often trained on vast amounts of text data from various web resources, including websites, books, and articles, which may not always be accurate, up-to-date, or relevant to plastic surgery. This lack of control over the training data can lead to the generation of information that is not evidence-based, inconsistent with current best practices, and at worst, nonsensical. By developing our own plastic surgery-specific LLMs, we can ensure the reliability of the generated content for clinical use by curating high-quality, domain-specific training data. Furthermore, the complex nature of decision-making processes in plastic surgery requires a deep understanding of specialty-specific factors that commercially available LLMs may not capture.3Buzzaccarini G. Degliuomini R.S. Borin M. et al.The Promise and Pitfalls of AI-Generated Anatomical Images: Evaluating Midjourney for Aesthetic Surgery Applications.Aesthetic Plast Surg. 2024; (Published online January 18) (Published online January 18)https://doi.org/10.1007/s00266-023-03826-wCrossref PubMed Scopus (0) Google Scholar This is where the role of expert input becomes crucial. By involving plastic surgeons in the fine-tuning process, we can ensure that the LLMs not only learn from the available literature but also incorporate the clinical decision-making skills that come with years of practice. For instance, in a microsurgical DIEP flap for breast reconstruction, a plastic surgeon must consider various factors such as perforator location, quality, and optimal flap design. A specialty-specific multimodal LLM trained on microsurgery literature and expert insights could provide valuable recommendations based on similar cases and best practices. In addition, specialty-specific LLMs can enhance plastic surgery education and resident training.4Mohapatra D.P. Thiruvoth F.M. Tripathy S. et al.Leveraging Large Language Models (LLM) for the Plastic Surgery Resident Training: Do They Have a Role?.Indian J Plast Surg. 2023; 56: 413-420https://doi.org/10.1055/s-0043-1772704Crossref PubMed Scopus (2) Google Scholar Interactive learning platforms powered by these models can provide personalized feedback, simulate complex case scenarios, and facilitate self-directed learning. An LLM-powered patient simulator, for example, could allow residents to sharpen decision-making skills in a virtual clinical environment, thereby enhancing patient safety and the learning experience. To harness the full potential of LLMs in plastic surgery, collaboration and investment from our community are essential.5Spoer D.L. Kiene J.M. Dekker P.K. et al.A Systematic Review of Artificial Intelligence Applications in Plastic Surgery: Looking to the Future.Plast Reconstr Surg Glob Open. 2022; 10e4608https://doi.org/10.1097/GOX.0000000000004608Crossref PubMed Scopus (8) Google Scholar We should work together to curate high-quality datasets, establish standardized clinical guidelines for AI training, and develop robust validation frameworks to ensure the reliability and trustworthiness of our specialty-specific LLMs. Generative AI systems like LLMs have sparked tremendous excitement over their prospective value in healthcare training and delivery. Realizing the promise of this technology in plastic surgery requires a commitment to creating domain-specific LLMs tailored to the unique needs and challenges of our specialty. By investing time and energy in the development of specialty-specific models trained on carefully curated plastic surgery literature, guidelines and clinical datasets, and fine-tuned with expert input, we can create powerful and practical AI tools. With this enabling technology there lies vast potential to enhance clinical care and surgical outcomes, improve education and training, and drive research and innovation for our patients. Not required The authors have no financial interests to declare in relation to this article's content. None

Ähnliche Arbeiten