Artificial intelligence in radial diagnostics of breast cancer

Abstract

Breast cancer (BC) ranks second in prevalence among all malignant tumors and is the most frequent cancer in women. This literature review details the introduction of artificial intelligence (AI) systems based on ultra-precise neural networks into clinical practice. This direction in diagnostic medicine is very promising, and in many ways can improve the existing and firmly entered into everyday practice methods of breast imaging. Such methods include mammography, ultrasound (USG) and magnetic resonance imaging (MRI). Mammography screening is an advanced tool for early detection of breast cancer, which has reduced the mortality rate from the disease by 30% in the last thirty years. Nevertheless, the method has potential drawbacks, including false positives and false negatives due to the phenomenon of tissue summation on a two-dimensional image, as well as the increased density of anatomical structures of the breast. Artificial intelligence systems are designed to improve and simplify this imaging method, reducing the time required for image interpretation. At present, these digital systems for their implementation in practice are not yet sufficiently studied, there are many errors and flaws in the interpretation of mammograms. The next widely used method of breast visualization is ultrasound. This method is able to detect neoplasms hidden by mammography in women with anatomically dense breast tissue structure, which makes it particularly useful in cancer diagnosis in women of reproductive age. However, ultrasound also has its disadvantages, among which stand out the operator-dependence of the method. Currently, the artificial intelligence system S-detect (2018, Samsung Medison) is actively used, which is able to interpret the image, automatically reading information in real time, thus increasing the effectiveness of ultrasound. The program has already demonstrated high sensitivity, specificity and accuracy (95.8%, 93.8%, 89.6%, respectively) in detecting benign and malignant breast masses in the trial phases from 2019. To date, there are no approved guidelines for the use of artificial intelligence programs in ultrasound diagnostics, with further research and evidence of the utility of such synergy required. Artificial intelligence programs combined with MRI diagnostics have also demonstrated increased efficiency and sensitivity of the method. However, false positives and false negatives (including missed pathology) have also been reported in this combination.
 A literature review of PubMed and Google Scholar article databases was performed. The focus was on full-text articles.