Performance of a Machine Learning Model Based on Routinely Collected Radiological and Clinical Data in Predicting Axillary Metastasis in cN0 Early Stage Breast Cancer

Abstract

Objective: Axillary lymph node metastasis (ALNM) is one of the main factors determining treatment and prognosis in breast cancer. This study presents a machine learning (ML) model for the prediction of pretreatment ALNM based on tru-cut biopsy results and the radiological characteristics of the tumor.

Methods: The mammography (MG) and ultrasound (US) findings of patients diagnosed with invasive breast cancer who underwent sentinel lymph node biopsy (SLNB) between 2017 and 2024 were examined, along with the histopathological characteristics of the tumors.

Results: A total of 361 patients were included in the study, with a median age of 52 years (43-64 years) and median lesion size of 24 mm (16-36 mm). While 52.9% of the patients (191) had no ALNM, 47.1% (170) presented with metastasis. Among the evaluated models, LightGBM (AUC: 0.73, accuracy: 0.68) and XGBoost (AUC: 0.734, accuracy: 0.654) successfully predicted axillary metastasis.

Conclusion: The ML model developed using MG, US, and tru-cut biopsy results commonly used in daily practice successfully predicted ALNM.

Cite this article as: Rona G, Kökten Ş, Serel TA, Yıldız Ü, Baysal T. Performance of a machine learning model based on routinely collected radiological and clinical data in predicting axillary metastasis in cN0 early stage breast cancer. Cerrahpaşa Med J. 2026; 50: 0100. doi: 10.5152/cjm.2026.25100.