OBJECTIVES: Breast cancer is one of the global health concerns and a leading cause of death in the female population. It demands robust early diagnostic strategies for enhanced recovery rates and diminished mortality. This study aims to elevate breast cancer classification by integrating feature and histological imaging data through advanced computational methods.

METHODS: A comparative analysis of six machine learning (ML) algorithms was conducted – Logistic Regression, Support Vector Machine (SVM), Decision Tree, Random Forest, Gaussian Naïve Bayes, and Linear Discriminant Analysis. The Wisconsin Diagnostic Breast Cancer (WDBC) dataset was utilized. Furthermore, the capabilities of the four convolutional neural network (CNN) architectures (SqueezeNet, GoogleNet, ResNet-50, and DenseNet-121) were employed on the BreakHis histological imaging dataset. Transfer learning techniques and oversampling were employed to address data imbalances. Evaluation metrics such as the area under the receiver operating characteristic curve (AUC), accuracy, precision, sensitivity, specificity, negative predictive value, and F1-score were used, with consistent settings facilitated by MATLAB R2023b (The MathWorks, Inc.).

RESULTS: The feature dataset analysis shows SVM with a superior classification performance with an AUC of 0.999 and an accuracy of 98.24%, followed by Logistic Regression and Random Forest, 97.64% and 95.88%, respectively. Regarding the histological imaging data, DenseNet-121 with 400x and 200x showed significant performance, while ResNet-50 outperformed at lower resolutions (100x and 40x). SqueezeNet, GoogleNet, ResNet-50, and DenseNet-121 exhibited average accuracies of 87.94%, 94.51%, 96.92%, and 96.94%, respectively.

CONCLUSIONS: Matlab-pretrained models exhibited unparalleled performance compared to previous studies on alternative platforms. The utilization of diagnostic imaging remains pivotal in identifying breast cancer, underscoring the significance of these findings. The results advocate for integrating machine learning algorithms and histological imaging, providing a comprehensive and practical approach to breast cancer classification.