Heuristic Neural Network for Thermography Breast Cancer Detection

Thermography, being low cost and radiation free, can be the most accessible tool for early detection of breast cancer through mass scanning. But it’s accuracy needs to be improved. Currently, the state-of-the-art classification method in use with thermographic breast cancer detection algorithms serves as a bottleneck due to their limited accuracy and applicability to only a subset of available thermographic breast images with some preconditions. So, a more accurate algorithm applicable to all thermographic breast images without manual intervention is a current need. This article proposes a heuristic algorithm for machine learning algorithms for detecting thermogram breast cancer. The pre-processing of the input thermogram image is accomplished by contrast enhancement, and mean filtering. Then the Gradient Vector Flow Snakes (GVFS) is adopted for breast segmentation. From the segmented images, the entropy-based features are extracted. In the classification phase, a Heuristic-based Neural Network (HNN) is introduced, which diagnoses the breast cancer-affected images. The modification on NN is extended by the Oppositional Improvement-based Tunicate Swarm Algorithm (OI-TSA). Using a benchmark database, the proposed CAD system was evaluated based on accuracy and different performance matrices. The analysis and experimental results showed that our system would contribute to the promising future in the case of breast cancer detection using thermography.