Non-Invasive Logistic Regression Model for Predicting Regional Lymph Node Metastases in Cutaneous Melanoma

Authors:

Abstract
Relevance : Cutaneous melanoma is characterized by high metastatic potential, and regional lymph node involvement is a key prognostic factor. Purpose : To develop and evaluate a non-invasive logistic regression model for predicting regional lymph node metastases. Materials and methods : A retrospective single-center study included 121 patients. Clinical, ultrasound, elastography and scintigraphy parameters were used as predictors; morphological data were excluded. Results : The model demonstrated a high discriminative ability (ROC AUC ≈ 0.94). At a threshold of p = 0.316, the sensitivity was ≈0.85, the specificity was ≈0.96, the accuracy was ≈0.93, at a lower threshold of p = 0.103, the sensitivity reached about 0.93 with a specificity of about 0.79. Conclusion : The model may be used for primary risk stratification and guiding invasive staging decisions.

This publication has 5 references indexed in Scilit: