Investigations using deep learning algorithms for breast tumor diagnosis

Abstract

Across globally, breast cancer is the leading cause of mortality newlineamong women which accounts approximately 12% of the new cases. For past newlinefive years, nearly 7.8 million women are diagnosed with breast cancer in newlineIndia. According to the Indian Council of Medical Research (ICMR), the newlinebreast cancer is the most prevalent cancer as there is a high incidence in both newlinerural and urban cities. The mortality rate due to this disease can be reduced by newlinediagnosing the disease at the earliest. This is achieved through several newlineimaging modalities such as a regular screening or follow ups. The preliminary newlinetest for disease diagnosis is digital mammography, where the analysis of the newlinedisease is carried out using Artificial Intelligence (AI) based techniques. newlineDespite the great success of Computer Aided Diagnosis (CAD), there are newlineseveral other challenges involved in diagnosis such as women with high dense newlinebreast newlinetissues, newlinemorphological characteristics of the masses and newlinemicrocalcification. Hence, assisting the radiologists and other clinicians with newlinean efficient CAD system is a substantial goal in medical image analysis. newlineRadiologists can utilize them as an assisting tool in detection and newlineclassification of tumor lesions. newlineThe amelioration of computational systems along with the deep newlinelearning techniques has revolutionized breast tumor diagnosis by significantly newlineincreasing the accuracy, precision, efficiency, and predictions. Amongst newlineseveral challenges, access to high quality data for training and validating the newlinemodels encompasses several key issues such as heterogeneity, newlinestandardization, bias in the dataset, labelling and annotations. newline