CT-based texture analysis for preoperative prediction of key histopathological features in laryngeal cancer

Purpose: Laryngeal squamous cell carcinoma (LSCC) is a malignancy with significant morbidity and mortality. Accurate preoperative assessment of key histopathological features, including lymph node metastasis (LNM), extranodal extension (ENE), lymphovascular invasion (LVI), perineural invasion (PNI), and thyroid cartilage invasion (TCI), is crucial for optimising treatment strategies. Computed tomography (CT)-based texture analysis, a radiomics approach, has shown potential in identifying tumour heterogeneity and aggressive histopathological behaviour. The aim of the study was to evaluate the predictive value of CT-based texture analysis for detecting key histopathological features in LSCC and assess its potential as a non-invasive tool for preoperative risk stratification. Material and methods: tests and receiver operating characteristic (ROC) curve analyses were performed to assess diagnostic performance. Results: < 0.05). GLZLM_HGZE ≤ 4635 predicted LNM with an AUC of 0.847 (95% CI: 0.708-0.986), sensitivity of 71%, and specificity of 84%. For ENE, GLZLM_HGZE ≤ 4625 achieved an AUC of 0.891, sensitivity of 85%, and specificity of 84%. TCI prediction was highest with GLZLM_SZLGE ≥ 0.00118 (AUC: 0.964, 95% CI: 0.909-1.000), with sensitivity of 88% and specificity of 93%. No significant predictors were found for LVI or PNI. Conclusions: CT-based texture analysis is a promising non-invasive tool for preoperative risk assessment in LSCC, particularly for LNM, ENE, and TCI. Further validation in larger studies is warranted.