Magnetic resonance imaging (MRI) is widely used to diagnose breast cancer. However, the high sensitivity of dynamic contrast-enhanced MRI (DCE-MRI) comes at the cost of increased complexity in image interpretation and segmentation, which requires advanced techniques to help radiologists delineate tumor boundaries accurately and efficiently. This work proposes an automated methodology that reduces the reliance on manual annotations by integrating unsupervised clustering with supervised classification to identify the tumor cluster within masks derived from clustering techniques using only morphological characteristics, thus facilitating rapid, accurate, and reproducible tumor detection in breast magnetic resonance imaging. The approach consists of an initial segmentation using Fuzzy C-Means followed by classification using supervised machine learning models. A comprehensive feature selection analysis that identifies the most informative descriptors for tumor classification, resulting in a classification accuracy of 99.44%, with consistent improvements across all evaluated metrics. The proposed approach mitigates annotation dependency while ensuring interpretability, representing a practical tool for clinical deployment.