Leveraging Artificial Intelligence for Effective Breast Cancer Management in Low-Resource Centers: A Pilot Project of Tnm Staging Machine

Leveraging artificial intelligence in breast cancer management offers opportunities to address a major public health challenge in low-resource settings. Yet the challenges of implementing AI in developing regions may contribute to global disparities due to the considerable cost of infrastructure, data acquisition, data cleaning, and modeling. Aim: To lay the foundation for AI integration into a clinical workflow in a low-resource center by training and evaluating a machine learning model for breast cancer staging. Methods: In a Python Jupyter notebook environment, using Pandas, Numpy, and Sklearn (Scikit-learn), we trained a decision tree classifier (DTC) model to learn the rules of the 8th edition of the AJCC Breast Cancer Staging for Breast Cancer. The model was trained on 158 samples derived from the AJCC 8th Edition staging guidelines and validated via 39 real-world anonymized records. The performance evaluation was performed with metrics derived from a confusion matrix and 3-fold cross validation. Result: Starting with a high dissimilarity in the global datasets with a Gini impurity of 0.807 in the input data, the model achieved pure classifications with a Gini impurity of 0.0, indicating no misclassifications. The decision tree classifier achieved 100% accuracy, precision, and recall on all the test and real-world datasets. Conclusion: We trained a decision classifier AI model to learn and implement the AJCC breast cancer staging guidelines with perfect performance, demonstrating the feasibility of cost-effective AI solutions for breast cancer staging in low-resource settings and paving way for broader clinical integration, such as integration, into treatment recommendation systems.