Comparative analysis of diagnostic ultrasound and histopathology for detecting cervical lymph node metastases in head and neck cancer

Abstract

Purpose: We evaluated the current performance of diagnostic ultrasound (US) for detecting cervical lymph node (LN) metastases based on objective measures and subjective findings in comparison to the gold standard, histopathological evaluation.

Patients and methods: From 2007 to 2016, we prospectively included patients with head and neck cancer who were scheduled for surgical therapy including neck dissection. LNs were examined by multimodal US by a level III head and neck sonologist and individually assigned to a map containing six AAO-HNS neck LN levels preoperatively. During the operation, LNs were dissected and then assessed by routine histopathology, with 86% of them examined individually and the remaining LNs (14%) per AAO-HNS neck LN level. The optimal cutoff points (OCPs) of four defined LN diameters and 2D and 3D roundness indices per AAO-HNS neck LN level were determined.

Results: In total, 235 patients were included, and 4539 LNs were analyzed by US, 7237 by histopathology and 2684 by both methods. Of these, 259 (9.65%) were classified as suspicious for metastasis by US, whereas 299 (11.14%) were found to be positive by histopathology. Subjective US sensitivity and specificity were 0.79 and 0.99, respectively. The OCPs of the individual LN diameters and the 2D and 3D roundness index were determined individually for all AAO-HNS neck LN levels. Across all levels, the OCP for the 2D index was 1.79 and the 3D index was 14.97. The predictive performance of all distances, indices, and subjective findings improved with increasing metastasis size. Anticipation of pN stage was best achieved with subjective US findings and the smallest diameter (Cohen's κ = 0.713 and 0.438, respectively).

Conclusion: Our LN mapping and meticulous 1:1 node-by-node comparison reveals the usefulness of US for detecting metastatic involvement of neck LNs in head and neck carcinomas as compared to histopathology. The predictive ability for small tumor deposits less than 8 mm in size remains weak and urgently needs improvement.

Keywords: Diagnostic ultrasound; Lymph node metastasis; Roundness.

Fig. 1

Schematic 3D representation of a bean-shaped LN. The distances measured by ultrasound were as follows: Longest Cross-Sectional Diameter (l), Longest Perpendicular Diameter (b), Longest Diameter Perpendicular to the Plane (d1), Longest Diameter Perpendicular to d1 (d2). A spherical, concentrically growing mass (metastasis) will initially lead to an increase in the smaller distances b and d2 and only secondarily to the distances d1 and l