Radiologist beats computers one to zero. In a nutshell, this is the message that emerges from a large study published on the pages of British Medical Journal regarding the ability to identify potential breast cancers through images obtained from mammography screening. But be careful to think that the game is over. Artificial intelligence does not go into the attic at all and can already be used successfully today to help doctors identify possible “areas” that deserve further investigation.
Early diagnosis and mammography screening
In all cancers, especially those of the breast, early detection can make a real difference. The sooner the disease is diagnosed, the greater the chances of success. For example, when the tumor is identified in its earliest stages, ie when we are faced with an “in situ” neoplasm, survival 5 years after diagnosis is 98%. But in order to achieve this extraordinary result, periodic checks are essential. Control that also passes from mammography screening offered free of charge by our national health system to all women aged between 50 and 69 years – even if there are regional differences in the age range, often anticipated and extended by 5 years (therefore from 45 to 75) The purpose of screening – and mammography – is to offer the maximum possible protection free of charge to citizens with the limited resources available. For this reason, mammography screening focuses on women between 50 and 69 years of age, the age group in which breast tumors are more frequent and mammography is more effective.
Read mammograms
Through mammography, the doctor is enabled to visualize the structure of the breast. A sort of radiography useful for directing the diagnosis but which is difficult to distinguish between the different types of formations. For this reason, the examination gives a general indication to carry out further investigations. Over the years, thanks to the development of artificial intelligence, software has been developed that can read the images of mammograms in order to identify any lesions to be investigated. At present, however, the use of software alone is not always suitable for reading images from screening. In the analysis published by British Medical Journal, the result of the review of 12 papers comparing the ability of the software to detect cancerous lesions with the capabilities of the radiologist, it was found that in the three largest retrospective studies, which included nearly 80,000 women screened in Europe and the United States, most some of the AI systems (34 out of 36, 94%) were less accurate than a single radiologist and were all less accurate than the screening of two or more radiologists. accuracy of artificial intelligence in breast cancer screening programs, and it’s unclear where it could be used to help. A complementary use to the work of radiologists, rather than replacing them, could be a promising possibility, perhaps using artificial intelligence to pre-select normal and easy mammograms that do not need further revisions ”.
Present and future applications
But while waiting for artificial intelligence to replicate the performance of the eye of an expert radiologist, there are already interesting and valid applications on its use: with artificial intelligence it is beginning to be possible to identify potential pathologies that in in fact, they did not have to be examined by making a specific diagnostic test. An example? Through artificial intelligence it is now possible to obtain some information from the image such as the presence of bone fragility and osteoporosis risk, the accumulation of calcium in the coronary arteries, microfractures and fatty liver. Useful information to be able to intercept the signs of a potential pathology very early.
.