Related Experiment Video
Updated: Jun 7, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.7K
Explainable machine learning versus known nomogram for predicting non-sentinel lymph node metastases in breast cancer
Asieh Sadat Fattahi1, Maryam Hoseini2, Toktam Dehghani3
1Department of Surgery, Faculty of Medicine, Endoscopic and Minimally Invasive Research Center, Mashhad University of Medical Sciences, Mashhad, Iran.
Computers in Biology and Medicine
|November 17, 2024
Summary
Machine learning models, particularly Random Forest, show improved accuracy in predicting non-sentinel lymph node metastasis in Iranian breast cancer patients compared to the MSKCC nomogram. This enhances personalized treatment strategies.
Area of Science:
- Oncology
- Medical Informatics
- Artificial Intelligence in Medicine
Background:
- Axillary lymph node dissection (ALND) is standard for breast cancer with positive sentinel lymph nodes (SLN).
- Predicting non-sentinel lymph node (NSLN) metastasis is challenging, leading to potential overtreatment with ALND.
- The MSKCC nomogram aids prediction but requires validation in diverse populations like Iranian breast cancer patients.
Purpose of the Study:
- To compare the predictive performance of the MSKCC nomogram against various machine learning (ML) models for NSLN metastasis.
- To evaluate the utility of eXplainable Artificial Intelligence (XAI) in identifying key predictors of NSLN metastasis.
- To assess the applicability of these prognostic tools in the Iranian breast cancer population.
Main Methods:
- Comparative analysis of the MSKCC nomogram and ML models (Random Forest, Logistic Regression) using a dataset of 183 Iranian breast cancer patients.
- Utilized 16 clinical features and employed eXplainable Artificial Intelligence (XAI) methodologies, including SHAP values, for model interpretability.
- Rigorous statistical evaluations, including training and validation of ML models, were performed.
Main Results:
- The Random Forest (RF) model achieved higher accuracy (72.2%) and AUC (0.77) than the MSKCC nomogram (AUC 0.73).
- Logistic Regression (LR) also showed competitive performance (accuracy 65%, AUC 0.73).
- RF model demonstrated high sensitivity (75%) and precision (73%), identifying ductal carcinoma in situ (DCIS) and tumor characteristics as critical predictors. XAI provided feature importance insights.
Conclusions:
- Machine learning models, especially Random Forest, offer superior prediction of NSLN metastasis compared to the MSKCC nomogram in Iranian breast cancer patients.
- Findings support the development of tailored prognostic tools for diverse populations to guide personalized treatment decisions.
- This research can improve the accuracy and efficacy of breast cancer management in the Iranian healthcare system.
