<![CDATA[Early detection of breast cancer is crucial to improve survival rates and reduce mortality. However, the process of classifying histopathological images often faces challenges due to limitations in visual features and data imbalance between classes. This study aims to improve the accuracy of breast cancer image classification by combining texture feature extraction using Gabor Filters and a Convolutional Neural Network (CNN) classification model. The dataset consists of 50×50 pixel images extracted using Gabor filters with varying orientations and frequencies to capture tissue texture characteristics. To address class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was used prior to model training. The CNN model was trained using a batch size of 32 and optimized using the Adam algorithm. The evaluation results show that the model is capable of achieving an accuracy of 97.06%, precision of 81.25%, recall of 86.67%, and an F1-score of 83.87%. The high recall and F1-score values indicate the model's ability to detect cancer cases effectively and evenly. Thus, the combination of Gabor Filter, SMOTE, and CNN has great potential to be implemented as a breast cancer diagnosis support system based on histopathological images.]]>
Copyrights © 2026
