Breast cancer affected approximately 66 out of every 100,000 women between 2023 and 2025, making it the most common type of cancer among women worldwide. Early detection remains a crucial factor for successful treatment and improved patient outcomes. Among the various diagnostic techniques, histopathological imaging plays a fundamental role in the accurate identification and classification of breast cancer. In this context, the objective of this study was to develop a novel segmentation method tailored for histopathological tissue analysis, focusing on structures such as tumors, stroma, lymphocytes, and other relevant components. The proposed approach, named Early Learning, was designed to enhance the training process of deep neural networks by introducing segmentation feedback at early layers of the architecture. This method aims to increase training efficiency and segmentation precision. Experimental results demonstrated promising performance, achieving an Intersection over Union (IoU) of 0.6703 and a Dice Similarity Coefficient (DSC) of 0.7099. The integration of Early Learning into the U-Net architecture proved effective, reinforcing its potential to assist in breast cancer diagnosis and support histopathological decision-making.