Implementation of machine learning models for invasive Breast cancer classification – towards early diagnosis

Abstract

Invasive lobular carcinoma (ILC) and invasive ductal carcinoma (IDC) are the most common forms of breast cancer and leading causes of death among women globally. Early detection of these subtypes with the aid of routinely collected demographic and clinical data is a feasible strategy and facilitates cancer routine care. Due to its widespread use, it could potentially benefit from the use of machine learning models to aid clinicians in their diagnosis workflow. Thus, we aimed to develop ML models using data retrieved from electronic health records corresponding to 1021 patients to accurately predict and classify these subtypes. Of note, prediction models were built using decision trees, naïve bayes, logistic regression, random forest and support vector machine. The classifiers' performances were assessed using accuracy, precision, recall and area under the curve. It is worth noting that the SVM model outperformed the ILC and IDC diagnostic response with an overall accuracy of 83% (AUC score 85%, precision 88% and recall 30%) than the other models. These findings highlight that it is feasible to develop a precise prognostic model by fusing various clinical and demographic data. The positive outcomes show the promise of using machine models as diagnostic aid tools for pathologists in clinical practice.