This study examines the integration of a knowledge-based Convolutional Neural Network (CNN) model for breast cancer histopathology image classification through ontological and epistemological perspectives. Ontologically, the research focuses on the digital representation of histopathological breast tissue images as entities representing benign and malignant conditions, establishing a stable and comprehensive mapping of tissue morphological characteristics. Epistemologically, the study employs a deep learning approach using a CNN model to acquire and validate knowledge about cancer cell morphology patterns from image data, constructing robust epistemic claims regarding tissue differentiation. The BreakHis dataset comprises 7,909 images resized to 224×224 pixels that underwent preprocessing normalization and image augmentation to enhance data quality. The CNN model was designed with Adam and SAM optimizers, learning rates of 0.0001 and 0.003, and a three-epoch warm-up phase to maintain training stability. Experimental results achieved training accuracy of 0.8432, testing accuracy of 0.8481, AUC of 0.8318, precision of 0.8124, and recall of 0.8966, demonstrating excellent model performance in recognizing cancer tissue patterns without overfitting. The integration of this knowledge-based CNN model contributes theoretically to the advancement of artificial intelligence and biomedical science, while demonstrating practical relevance as a reliable decision-support system for breast cancer diagnosis.