Robotic thyroidectomy using Artificial Intelligence (AI) model in real-time recognizing Recurrent Laryngeal Nerves (RLN)

Abstract

Background

Robotic surgery is gaining momentum in thyroidectomy particularly among young female patients. Swallowing, voice disorders, throat discomfort and cervical scarring represent major sources of anxiety due to social and occupational concerns. This has driven demand for precise and scarless approaches to improve safe and cosmetic outcomes to reduce psychological burden [1]. While endoscopic thyroidectomy has gained popularity, its technical limitations including restricted endoscopic visualization, limitation of fixed magnification, and the ‘chopstick' effect of rigid instruments compromise on precision and safety during dissection along RLN. Robotic thyroidectomy emerges as a promising alternative, offering three-dimensional high-definition images, magnification, multiple forceps articulation, tremor-stabilization function and motion scale function [2]. Concurrently, advances in AI-powered intraoperative image recognition are enhancing real-time identification of critical structures like the RLN during procedures, further improving surgical precision and safety [3,4].

Methods

The video demonstrates robotic thyroidectomy in a 42-year-old female patient with a 0.6 × 0.5 × 0.5cm irregularly shaped nodule (TI-RADS 4a) at the dorsal mid-portion of the right lobe, presenting ill-defined margins, taller-than-wide sign, and capsular invasion with multiple abnormal right central lymph nodes. Ultrasound-guided fine-needle aspiration (FNA) confirmed papillary thyroid carcinoma (Bethesda Category VI) with BRAF V600E mutation positivity. Given her strong preference against cervical scarring and expressed concerns regarding postoperative voice changes, she underwent robotic thyroidectomy with a novel AI-assisted real-time RLN navigation.

Results

We performed right lobectomy with isthmusectomy and concomitant right central compartment lymph nodes dissection (CCND). The AI system provided continuous real-time guidance with robust performance across diverse surgical conditions (Fig. 1). Total operative time was 120 minutes with minimal blood loss. The patient experienced no postoperative hoarseness and discharged on the third postoperative day. Final pathology confirmed papillary thyroid carcinoma (classical variant) with metastatic involvement in perithyroidal and right paratracheal lymph nodes (2/5).

Conclusion

This video demonstrates the feasibility and safety of AI-assisted real-time recognition in dissection along RLN during robotic thyroidectomy for complex cases. This approach pioneers a robotic platform plus AI real-time navigation paradigm in conjunction with intraoperative nerve monitoring (IONM) to assist accurate identification and functional confirmation of the RLN, standardizing surgical procedures across institutions and surgeons, alleviating the fatigue and anxiety caused to the surgeon due to continuously monitoring the neural structure during the operation and shortening the learning curve for trained surgeons. It represents a pivotal transition from experience-dependent surgery to data-driven intelligent surgery, significantly advancing robotic surgery.