Texture Features-Based Breast Cancer Detection Using Artificial Neural Network

Globally, the incidence rate of breast cancer is very high. Substantial support for awareness of breast cancer and research plays a vital role in diagnosing and treating the disease earlier, before it becomes harmful. It has been the focus point to detect breast cancer with many features space without concern about computational time, speed, and complexity of an algorithm. Instead of adopting a complex procedure, an easy approach is considered which not only detects cancer in less time but also reduces computational time and complexity; Such a model can be achieved by reducing features space and steps that can lead to accurate detection. Three steps-based algorithms, namely features extraction, features selection, and classification, are used to distinguish mammograms. Grey Level Co-occurrence Matrix (GLCM) was used for features extraction in which a short number of texture features subset (10 features) is considered in this work. This was further reduced to six features by using relief algorithm proven to be a computationally efficient and capable of detecting features dependencies. At last, Feedforward Neural Network (FNN) is used for classifying the data. This approach is easy to implement and fast in detection without defacing results. The proposed system shows 99.8% sensitivity, 100% specificity, and 99.9% accuracy.