Review article
https://doi.org/10.55378/rv.46.1.1
Application of CAD in the diagnosis of breast cancer
Lucija Bratinčević
; Department of Health Sciences University of Split, Split, Croatia
Tatjana Matijaš
orcid.org/0000-0003-0494-1601
; Department of Health Sciences University of Split, Split, Croatia
Abstract
Breast cancer is an extremely dangerous disease, which if diagnosed in time has a high survival rate. The incidence and mortality rate from breast malignancies are increasing, so in order to reduce these numbers, new technological solutions are being looked into, that should enable the earliest possible detection of breast carcinoma. Although the original solution was seen in traditional computer-assisted detection systems, CAD, applied to various radiological breast imaging methods, the results of various studies discussed in this paper found that they did not meet their original expectations in breast cancer diagnosis. The use of conventional CAD systems still had too many limitations such as a decrease in particularity and positive predictive value with an increase in incorrect results and an increase in revocation rates. However, the development of artificial intelligence, AI-based algorithms has improved the quality and accuracy of conventional CAD systems. Unlike conventional CAD systems based on hand-crafted features, deep learning, as a subfield of AI, is based on representational learning. In representational learning, the algorithm itself during training determines features in the image that indicate the presence of lesions. Recently, such deep learning algorithms have been applied to various radiological methods. In this paper, the analysis of various studies discusses the possibilities, but also the limitations of new AI-based applications for different modalities of breast imaging. Due to the small number of studies conducted on the topic of AI systems and the need for an extremely large set of data for training and validation of the algorithm, many scientists still doubt this new method. Despite these limitations, the AI approach has the ability to detect useful features in an image that are still invisible to the human eye. Future advances in technology will significantly improve AI systems and their implementation in health systems will be inevitable.
Keywords
AI, breast cancer, CAD, digital breast tomosynthesis, mammography, radiomics
Hrčak ID:
276116
URI
Publication date:
6.5.2022.
Visits: 1.049 *