Solmaz Sohrabei - Department of Development, Management and Resources; Office of Statistic and Information Technology Management, Zanjan University of Medical Sci- ences, Zanjan, Iran
Raheleh Salari - Poostchi Ophthalmology Research Center, Department of Ophthalmology, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
Seyed Mohammad Ayyoubzadeh - Department of Health Information Management, School of Allied Medical Sci- ences, Tehran University of Medical Science, Tehran, Iran
AlirezaAtashi Atashi - Department of E-Health, Virtual School, Tehran University of Medical Sci- ences, Tehran, Iran & Medical Informatics Research Group, Clinical Research Department, Breast Cancer Research Center, Motamed Cancer Institute (ACECR), Tehran, Iran

Introduction: one of the foremost usual methods for evaluating breast cancer is the removal of axillary lymph nodes (ALN) which include complications such as edema, limited hand movements, and lymph accumulation. Although studies have shown that the sentinel gland condition represents the axillary nodules context in the mammary gland, the efficacy, and safety of the guard node biopsy need to be evaluated. Subsequently, predicting axillary lymph node status before sentinel lymph node biopsy needs regular clinical data collection and would be supportive for oncologists and could keep the clinicians away from this strategy. Predictive modeling for lymph node statues may be one way to diminish the axillary lymph node dissection (ALND) and consequences. Methods: The database used in this study was provided by Clinical Research Department, Breast Cancer Research Center, Motamed Cancer Institute (ACECR), Tehran, Iran. It contains clinical and demographic risk factors records of ۵۱۴۲ breast cancer patients from which a total of ۳۸ features were selected. We performed modeling; based on six data mining algorithms (Decision Tree, Nave Bayesian, Random Forest, Support Vector Machine, Fast Large Margin, and Gradient Boosted Tree (GBT)). For evaluating the model, we used ۱۰-fold cross-validation in Rapid Miner v۹.۷.۰۰۱. Results: The results showed that the GBT model has a higher ability to predict lymph node metastasis than other models with an receiver operating characteristic (ROC) of ۹۷%, a sensitivity of ۹۶.۵۹%, an accuracy of ۹۰%, and specificity of ۸۱% Conclusions: Obviously, we have to diagnose cancer with a needle biopsy before surgery. Used data mining predictions and use of them to create a clinical decision support system for predicting cancer and lymph node statuses can help physicians and pathologists make the best decision for a patient's ALN surgery.

Lymph Nodes, Axilla, Decision Support Systems, Clinical, Machine Learning, Breast Neoplasms