Region-based feature enhancement using channel-wise attention for classification of breast histopathological images

Abstract Breast histopathological image analysis at 400x magnification is essential for the determination of malignant breast tumours. But manual these images tedious, subjective, error-prone and requires domain knowledge. To this end, computer-aided tools are gaining much attention in recent past as it aids pathologists save time. Furthermore, advances computational power have leveraged usage computer tools. Yet, to analyse challenging due various reasons such heterogeneity tumours, colour variations presence artefacts. Moreover, captured high resolutions which pose a major challenge designing deep learning models demands requirements. In context, present work proposes new approach efficiently effectively extract features from high-resolution images. addition, magnification, characteristics structure nuclei play prominent role decision malignancy. regard, study introduces novel CNN architecture called CWA-Net that uses channel module enhance potential regions interest nuclei. The developed model qualitatively quantitatively evaluated on private public datasets achieved an accuracy 0.95% 0.96%, respectively. experimental evaluation demonstrates proposed method outperforms state-of-the-art methods both datasets.