Background and Rationale NSCLC with supraclavicular or cervical lymph node metastasis (N3 disease) was historically considered unresectable. With the advance of multimodal therapy, complete lymph node dissection in selected patients has been shown to improve locoregional control. However, the supraclavicular fossa - frequently described as 'Pandora's Box' - is anatomically narrow and dense, with metastatic nodes often abutting the subclavian vessels, internal jugular vein, phrenic nerve, and brachial plexus. Conventional surgery relies heavily on the surgeon's tactile experience and 2D mental reconstruction of preoperative CT, increasing the risk of inadvertent neurovascular injury and incomplete clearance of post-treatment fibrotic versus active disease.
Intervention This study integrates two complementary technologies. (1) Multimodal image fusion using Fujifilm Synapse 3D maps PET SUV hotspots representing biologically active tumor onto a high-resolution CT anatomical model, producing a patient-specific digital twin. The fused model is exported to a static 3D format (OBJ/STL) and imported into the OpVerse XR platform - an offline data-conversion workflow with no real-time API coupling between the two systems, ensuring system stability and software compatibility. (2) Preoperative XR planning: in the operating room after general anesthesia and head/neck positioning, the surgeon wears an XR head-mounted display (HoloLens) and performs surface registration using stable bony landmarks. The 'see-through' overlay enables the surgeon to identify subcutaneous tumor hotspots and superficial venous anatomy, and to mark the optimal skin incision and initial dissection trajectory before any cut is made. The XR headset is then removed; lymph node dissection proceeds using standard sterile technique. Therefore the device functions purely as a preoperative visual aid (Non-Significant Risk), without entry into the sterile operative field.
Lymph node dissection definition En bloc systematic resection of the fibrofatty tissue containing metastatic nodes within the defined cervical or supraclavicular region, to achieve oncologic clearance and provide adequate tissue for pathologic staging and next-generation sequencing.
Statistical analysis All analyses are performed using SPSS. Continuous variables (operative time, blood loss, lymph node count) are summarized as mean ± SD or median; categorical variables (complications, registration success) as frequency and percentage. Successful completion rate of the XR-assisted workflow is reported with 95% confidence intervals. As a single-arm feasibility trial, no formal hypothesis testing is planned.
Preliminary data The investigators have completed a prior pilot trial (REC 202502149RINB, 2025) of XR-assisted lung nodule localization in 20 patients undergoing thoracoscopic sublobar resection, achieving 6-7 mm mean target registration error and 19/20 successful localizations without major complications. Because the supraclavicular region is more rigidly tethered to the bony skeleton and less affected by respiratory motion, registration accuracy in the present study is expected to be ≤5 mm.