Medical image based breast cancer diagnosis: State of the art and future directions

The intervention of medical imaging has significantly improved early diagnosis of breast cancer. Different radiological and microscopic imaging modalities are frequently utilized by medical practitioners for identification and categorization of different breast abnormalities by manual scrutiny. The meticulous classification of different breast abnormalities is challenging, because of ambiguous imaging data and due to indistinguishable characteristics of benign and malignant breast lesions. However, with the advent in applications of Artificial Intelligence (AI) in healthcare, researchers have turned their focus towards designing of efficient intelligent computer aided detection and diagnosis systems for prognosis of this catastrophic disease using image processing and computer vision (CV) techniques. An abundance of work could be found in literature on classification of different breast abnormalities, where majority of them has dealt with binary classification (i.e. benign and malignant). In current study, a comprehensive review has been presented to analyze and evaluate state of the art proposed methodologies for breast cancer diagnosis based over commonly used breast screening imaging modalities. The studies under consideration are mainly categorized into statistical machine learning based and deep learning based classifier, where deep classifiers further sub-categorized into models built from scratch and transfer learning based models. A number of factors have been taken to compare the performance of these classification models, on the basis of which some recommendations are provided for researcher to precede this work in future.

Automated summary

The intervention of medical imaging has greatly improved the early diagnosis of breast cancer, but the meticulous classification of breast abnormalities remains challenging. With the increasing application of Artificial Intelligence in healthcare, researchers are focusing on designing efficient intelligent computer aided detection and diagnosis systems to predict this catastrophic disease.