Deep learning assisting the surgical management of gynecologic cancers

This review examines the current role of deep learning in the surgical management of gynecologic cancers. It aims to evaluate applications across surgical training, intraoperative guidance, and outcome prediction, while identifying limitations that hinder translation into routine clinical practice.

Deep learning-based systems show promising performance in enhancing surgical skills through objective assessment and simulation platforms. Intraoperatively, deep learning models can recognize anatomical structures, surgical phases, and tumor tissue, with high diagnostic accuracy in selected settings. Emerging techniques such as surgical optomics and hyperspectral imaging may further improve tumor detection. Additionally, predictive models integrating clinical, imaging, and intraoperative data demonstrate potential in estimating complications, resource utilization, and survival outcomes. However, most studies are retrospective, based on limited datasets, and lack external validation, with minimal evidence of real-world clinical impact.

Artificial intelligence, particularly deep learning algorithms, represents a rapidly evolving tool in gynecologic oncologic surgery, with potential to standardize and personalize care. Nevertheless, significant gaps remain between experimental performance and clinical implementation.