Breast cancer detection and validation using dual modality imaging

Early detection of malignancy is important in order to reduce morbidity and mortality rate. However, existing approaches are based on a single modality having limited performance. Hence, the key objective of this study is detection and validation of breast cancer images using multimodality approach by combining both ultrasonography and mammography. Further, we investigate a particular segmentation approach that can accurately detect breast lesions using a dual-modality system. Three popular techniques i.e. Fuzzy-c-Means (FCM), and K-means (KM) and DPSO (Darwinian Particle swarm optimization) are experimented and evaluated for detecting lesions using combined ultrasonography and mammography. Performance measures such as the area of breast mass lesions, Jaccard Index (JI) and Dice Similarity Coefficient (DSC) have been used for the evaluation of the proposed dual modality breast cancer detection model. The results show that out of the implemented techniques, FCM outperformed others in detecting the lesion on both ultrasonography and mammography images with JI: highest average value (0.748) and lowest SD (0.124), DSC: highest average value (0.851) and lowest SD (0.084) and segmentation accuracy: highest average value (0.92) and lowest SD (0.074).