Enhancing Breast Tumor Classification in Microscopic Images through Fourier TransformBased High Pass Filtering

Анотація

Breast cancer remains a significant health concern globally, and early detection plays a pivotal role in improving patient outcomes. With the advent of digital pathology, computer vision techniques have emerged as powerful tools for automated analysis of microscopic breast tumor tissue images. The efficacy of Fourier Transform-based high pass filtering in enhancing the classification performance of a Convolutional Neural Network (CNN) model trained on microscopic breast tumor tissue images is investigated in this study. Initially, attempts to train the model on grayscale images yielded unsatisfactory results, indicating the limitations of conventional image representation for this task. However, notable improvements in classification accuracy were observed through the application of high pass filtering with varying radii, particularly with smaller radii. Furthermore, the analysis revealed a correlation between the radius of thehigh pass filter and model generalization, with larger radii leading to increased overfitting. These findings underscore the potential of Fourier Transform-based preprocessing techniques in augmenting the discriminatory power of CNN models.