Abstract:This article introduces a fast fractal method for classifying breast cancerous lesions in mammography. While fractal methods are valuable for extracting information, they often come with a high computational load and time consumption. This paper demonstrates that extracting optimal fractal information and focusing only on valuable information for classification not only improves computation speed and reduces process load but also enhances classification accuracy. To achieve this, we define an objective function based on accurate classification of benign and malignant masses to identify the best scale. Instead of extracting information from all nine scales, we extract and employ information solely from the best scale for classification. We validate the obtained scales using three classifiers: Support Vector Machine (SVM), Genetic Algorithm (GA), and Deep Learning (DL), which confirm the effectiveness of the proposed method. Comparative analysis with other studies reveals improved classification performance with the presented method.

An optimal fast fractal method for breast masses diagnosis using machine learning