Prospective Observational Study for Breast Microcalcifications' Classification With Artificial Intelligence Techniques

Breast microcalcifications are currently classified using the BI-RADS radiological scale. In case of suspicious microcalcifications (B3), it is recommended to perform a biopsy assessment for histopathological evaluation. However, about 70-80% of performed biopsies shows benign histology that does not require surgical treatment. Core biopsies are invasive procedures with a biological, psychological (patient discomfort), organizational and economic (for the Health Care System) costs. Therefore, accuracy's improvement in radiological classification of microcalcifications is essential. Recently, various approaches have been reported in the literature to detect and classify microcalcification as benign or suspicious in digital mammograms. Analysis methods based on the use of deep learning (DL) have also emerged as promising for processing mammography images. Convolutional neural networks (CNNs) are currently the state of the art for image classification in many application fields in the field of computer vision. This study intends to implement a classification method for breast microcalcifications (as benign or malign) with Artificial Intelligence (AI) techniques on mammographic images, evaluating the diagnostic performance (accuracy) of this approach. The evaluation will be conducted with reference to the standard radiological approach (BI-RADS classification).

Together with the application of AI systems to mammographic imaging, a further current clinical need is the development of a diagnostic tool able to determining in-situ the biomolecular characteristics of microcalcifications, accurately discriminating their nature without take tissue, fixation and embedding of the sample in paraffin, and without highly specialized evaluation by the pathologist. Raman spectroscopy (RS) is a highly specific method from the biomolecular point of view and, at the same time, it is compatible with in-vivo measurements. It consists in a biophotonic approach able to explore molecular composition of a given sample through its direct irradiation (through laser light) and the simultaneous acquisition of emission signals. RS information could be combined togheter with imaging features to implement an AI model for the combined classification of breast microcalcifications

Inclusion criteria

• Female subjects;
• Age between 18 and 88 years;
• Detection of microcalcifications on clinical and screening mammography with or without indication for histological assessment by biopsy;
• Subjects who agree to participate in the study by signing and dating the Informed Consent form

Exclusion criteria

• Personal history of breast cancer