AnisotropicBreast-ViT: Breast Cancer Classification in Ultrasound Images Using Anisotropic Filtering and Vision Transformer

Breast cancer classification in ultrasound images is a challenging and arduous task, primarily due to the quality of the images and the complexity of the lesions encountered. Early diagnosis of this pathology is known to increase patient survival chances. Therefore, computational methods have been developed to assist specialist physicians in this critical task. This study introduces AnisotropicBreast-ViT, a method that integrates anisotropic filtering, balanced data augmentation, and Vision Transformer to aid in the classification of breast ultrasound images. The proposed approach achieves promising results with an accuracy of 98.82%, specificity of 98.62%, sensitivity of 99.25%, precision of 97.08%, F1-score of 98.16%, and AUC-ROC of 0.989, surpassing current benchmarks in the field. These findings suggest that AnisotropicBreast-ViT has the potential to significantly improve breast cancer diagnosis, demonstrating its effectiveness and robustness in clinical applications.