A new study has found that artificial intelligence can be more accurate at diagnosing breast cancer than doctors, which could potentially pave the way for improved early detection.
At the moment, when women go for their routine mammogram, two radiologists will take a look at the scans to check for signs of cancer. If they disagree – which happens very rarely – a third doctor will assess the X-rays.
The new technology, pioneered by an international team of researchers, outperformed six radiologists in reading mammograms. The study involved the AI model being given anonymous X-ray images for analysis. Unlike the human experts, the AI model didn’t have any prior access to the patients’ medical records, so its findings were based solely on the mammograms.
The results of the study found that the AI was better at spotting cancer than a single doctor, and as effective as the NHS’s current double-reading system. Compared to one doctor assessing the scans, there was a 1.2 per cent reduction in false positives, which is when a mammogram is wrongly diagnosed as abnormal. There was also a 2.7 per cent drop in false negatives, where cancer is missed.
Radiology is a very specialised job, with radiologists undergoing years of medical training, but reading and interpreting X-rays is a time-consuming task – especially as the UK is currently experiencing a shortage of radiologists.
Although the AI system is yet to be tested in a clinic, it’s a positive step forward, as it’s a tool that could potentially be used alongside medical professionals to improve early detection and speed up the process. The technology isn’t designed to replace radiologists – at least one would still be in charge of diagnosis – but to improve and streamline the process.
Dr Mary Wilson is a Consultant Radiologist and Trustee for Prevent Breast Cancer. She shared her thoughts on the new findings: “The preliminary results from this study are very encouraging and given the severe lack of trained staff within breast imaging, it is possible such systems will be beneficial, though reading the mammograms is only a small part of the diagnostic process. We will now need to evaluate how such computerised systems might work within the UK Breast Screening Programme. However, any advance that helps to relieve the workforce pressures is to be welcomed.”