A ARTIFICIAL INTELLIGENCE AND BREAST RADIOLOGY
Cicero A. Urban · Mastology · 2019
A rtificial intelligence (AI) is a branch of computer science that researches the development of intelligent machines.Its current success results from a history of ups and downs.Since its development in the 1950s, it had moments of complete neglect, mainly in the decades of 1970 and 1980, with cuts in research funding due to discouraging initial results (called AI winter).However, its prestige started to improve at the end of the 1990s, especially after the Deep Blue computer, from IBM, defeated the world chess champion, Garry Kasparov, for the first time.In 2016, AI had another extraordinary victory.A neural network model called AlphaGo beat the world's greatest player of the board game Go, Lee Sedol.Currently, the progress of AI had an impact so large that its true history may be just beginning.The development of machine learning (ML) and deep learning (DL), the latter inspired in biology and mimicking the human cortex, made it possible to process a large volume of data and make complex inferences, often impossible for humans 1,2 .This technology is now reaching the medical field.Specifically in radiology, it can change the way exams are analyzed.Nonetheless, assuming that the role of AI would be restricted to this stage would be too naive.It has the potential to change the whole structure of a radiology clinic, from patient arrival to the delivery of results, reducing costs, and increasing agility 3 .Clinical practice has been implementing four fundamental systems in its procedures: • Lesion detection system: can identify and classify lesions with better performance than the traditional computer-aided detection (CAD); • Lesion quantification system: can quantify the lesion regarding its diameter, volume, and distance from anatomical structures (papillae, skin, and others), in addition to comparing the new exam with previous ones automatically; • Decision support system: helps to decide the best approach for the case, that is, it suggests an algorithm for research; • Differential diagnosis system: indicates the most likely diagnosis for the lesion, as well as the main differential diagnoses.