A new study has shown that AI-supported breast cancer screening is as safe and effective as double reading with two radiologists. It can cut doctors’ workload in half.
L’artificial intelligence (IA) is an invaluable ally in the analysis of the images of mammographythe main exam for diagnosis of the breast cancer. In fact, researchers have found that the screening mammografico conducted by one radiologist supported by one Properly trained AI has the same – if not greater – ability to identify i tumors from the “double reading”, the standard practice based on the survey of two radiologists. Simply put, artificial intelligence can speed up work and reduce the pressure on doctors, who are subjected to significant workloads due to the enormous number of x-rays to be evaluated. Due to the global shortage of radiologists, the support of AI is considered with great interest in this branch of medicine, however before introducing it into daily clinical practice it is essential to be sure of its work.
To determine the safety and theeffectiveness of AI, studies such as the recent one are needed MASAI (acronym of Mammography Screening with Artificial Intelligence), conducted by Swedish scientists from the Department of Translational Medicine of the University of Lund in close collaboration with colleagues from the Unilabs Mammography Unit – Skåne University Hospital and the Section for Breast Cancer Screening – Cancer Registry of Norway in Oslo (Norway). The researchers, coordinated by Professor Kristina Lång, professor at the Swedish University’s Division of Diagnostic Radiology, verified the effectiveness of AI with a randomized, controlled and population-based survey conducted on women aged between 40 and 80 years old. In simple words, they compared the diagnostic effectiveness of the classic double reading conducted by two radiologists with that of a single radiologist supported by AI, specifically the system called Transpara version 1.7.0.
In all, just over 80.000 donne with an average age of 54 years, whose mammograms – performed between April 12, 2021 and July 28, 2022 – were evaluated half by the radiologist and AI couple and the other half with the double standard reading by two radiologists. In all they were found 244 almost of breast cancer in group IA and 203 in the group evaluated by the two radiologists. Most interestingly, there hasn’t been an increase in false positives in the investigation assisted by artificial intelligence, that is, of mammograms evaluated as positive for cancer at the first check-up but which in reality are negative. “We found that using artificial intelligence resulted in 20 percent (41) more cancers being detected than standard screening, without affecting false positives. A false positive in screening occurs when a woman is recalled but cleared of cancer suspicion after screening.” Professor Lang said in a press release.
Drug mix blocks resistant breast cancer in lab tests: hopes for cure
Thanks to the support of the AI, the authors of the study explain, the workload on radiologists has been practically halved, with a reduction in readings on the screen of the 44 percent. A radiologist typically evaluates 50 images in an hour, Professor Lang points out: Group IA’s 40,000 mammograms took about five months less work to read than the other group. “Screening is complex. The balance between benefit and harm must always be taken into consideration. Just because a screening method finds more cancers doesn’t necessarily mean it’s a better method. What is important is to find a method that can identify clinically significant tumors at an early stage. However, this must be balanced against the harm of false positives and overdiagnosis of indolent tumours,” said Professor Lang.
It is still too early to incorporate AI into standard clinical practice, but it is clear that it is offering significant benefits at least at an experimental level. Recently, thanks to AI, a new antibiotic against a superbug; developed a unprecedented tumor analysis system by Italian researchers; and developed a blood test which detects 50 tumors with 75 percent accuracy. The details of the research “Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study” have been published in the authoritative scientific journal The Lancet.