The Diagnostic and Predictive Value of 18F-Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography in Laryngeal Squamous Cell Carcinoma

Abstract

This retrospective study examines the diagnostic accuracy of ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) and neck magnetic resonance imaging (MRI) in detecting nodal metastasis for patients with laryngeal squamous cell carcinoma (LSCC) and assesses the predictive values of metabolic and structural features derived from ¹⁸F-FDG PET/CT. By involving 66 patients from 2014 to 2021, the sensitivity and specificity of both modalities were calculated. ¹⁸F-FDG PET/CT outperforms neck MRI for nodal disease detection, with 89% sensitivity, 65% specificity, and 77% accuracy for nodal metastasis (p = 0.03). On the other hand, neck MRI had 66% sensitivity, 62% specificity, and 64% accuracy. Approximately 11% of patients witnessed a change in their therapy intent when relying on ¹⁸F-FDG PET/CT nodal staging results. Analyzing the cohort for PET-derived metabolic and morphological parameters, a total of 167 lymph nodes (LN) were visualized. Parameters such as the LN maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and LN size were computed. Logistic regression and receiver operating characteristic (ROC) analyses were performed. Among the 167 identified cervical LNs, 111 were histopathologically confirmed as positive. ROC analysis revealed the highest area under the curve for LN MTV (0.89; p < 0.01), followed by LN size (0.87; p < 0.01). Both MTV and LN size independently predicted LN metastasis through multivariate analysis. In addition, LN MTV can reliably predict false-positive LNs in preoperative staging, offering a promising imaging-based approach for further exploration.

Keywords: FDG PET/CT; MTV; TLG; laryngeal neoplasms; lymph nodes; lymphatic metastasis; squamous cell carcinoma.

Figure 1

(A) The receiver operating characteristic (ROC) analysis of lymph node (LN) factors, including LN maximum standardized uptake value (SUVmax), LN size, LN total lesion glycolysis (TLG), and LN metabolic tumor volume (MTV). (B) A maximum-intensity projection (MIP) image in a patient with histopathology-proven N0 laryngeal carcinoma demonstrated evidence of a large hypermetabolic mass confined to the larynx (arrowhead), in addition to a single mildly hypermetabolic right level II cervical LN (asterisk). (C) A coronal positron emission tomography/computed tomography (PET/CT) image revealed a cluster of 5 variably sized right cervical LNs (arrow), all of which were subcentimetric apart from the uppermost prominent LN (asterisk), measuring about 1 cm in the largest dimension and appearing mildly hypermetabolic (slightly above the liver SUVmax reference of 2.1). It is noteworthy that all visible LNs in this patient were below ROC cutoffs for size, SUVmax, MTV, and TLG. (D–F) MIP, axial PET, and axial PET/CT images of a patient with histopathology-proven N3 disease demonstrated evidence of bilateral hypermetabolic cervical lymphadenopathy (arrows), appearing in concordance with a large hypermetabolic laryngeal mass (arrowhead). All observed morphologic and metabolic metrics exceeded the optimal ROC cutoffs obtained from this cohort.

Figure 2

Scatterplots to visualize the correlations between the obtained factors. (A) LN TLG vs. LN MTV. (B) LN TLG vs. LN size. (C) LN TLG vs. LN SUVmax. (D) LN MTV vs. LN size. (E) LN MTV vs. LN SUVmax. (F) LN size vs. LN SUVmax.

Figure 3

Decision tree illustration of the predictive capability of LN MTV for LN false positivity. The analysis was performed by using the exhaustive Chi-squared Automatic Interaction Detector (CHAID) estimation procedure with a 10-fold cross-validation approach, and the results were corroborated using the Chi-squared Residual Tree (CRT) method.