Feasibility Study of Preoperative CT-Derived Volume and Intraoperative Tissue Mass Measurements for Lymph Node Yield Prediction in Cervical Lymphadenectomy.

Background: Standard treatment for oral cavity and oropharyngeal squamous cell carcinoma (SCC) includes cervical lymph node dissection (LND), where lymph node yield (LNY) is directly correlated with survival. Validated preoperative or intraoperative LNY prediction tools are lacking. This pilot study evaluates the feasibility of preoperative CT-derived volume estimates (CTV) of level II-IV neck dissection targets and direct intraoperative mass measurements (IOM) of resected lymphadenectomy specimens as potential tools for LNY prediction. Methods: 12 patients underwent standard-of-care level II-IV cervical LND by a single experienced surgeon with IOM recorded at closure. LNY was recorded using a standard LNY count and an additional comprehensive (CLNY) protocol. A novel segmentation protocol was designed to generate patient-specific level II-IV CTV estimates. Five independent readers tested protocol reliability using the Dice Similarity Coefficient (DSC) for segmentation overlap and the intraclass correlation coefficient (ICC) of CTV measurements. Pearson correlation coefficients (PCC) between IOM CTV and LNY metrics were reported. Results: This segmentation protocol showed high inter-rater reliability of CTV estimates (ICC = 0.85; 95% CI, 0.70-0.94) and substantial spatial overlap amongst readers (mean DSC = 0.72 ± 0.06). The strongest correlations, for which the study had adequate statistical power, included IOM with CLNY (Pearson correlation coefficient, PCC =0.87; 95% CI, 0.58-0.96) and mean CTV with IOM (PCC = 0.89; 95% CI, 0.65-0.97). Moderate correlations were observed between CTV and CLNY (PCC = 0.70; 95% CI, 0.20-0.91) and between IOM and standard LNY (PCC = 0.73; 95% CI, 0.26-0.92). Conclusions: The novel segmentation protocol produced reliable CTVs with strong correlation to intraoperative tissue-mass measurements (PCC = 0.89). While IOM showed strong correlation with CLNY (PCC = 0.87), both findings exceeded the threshold for adequate statistical power in this pilot study. The moderate correlations between CTV and LNY metrics (PCC = 0.59-0.70), though statistically significant, were below the study's power threshold and require validation in larger cohorts before conclusions can be drawn about their predictive utility.