An AI tool named Mia has shown its effectiveness in spotting cancer signs that were missed by human radiologists. This tool was tested alongside NHS clinicians in the UK, where it analyzed mammograms from over 10,000 women.
While the majority of these women were found to be cancer-free, Mia successfully identified all participants who showed symptoms of breast cancer. Impressively, it also detected 11 additional cases that the doctors had not noticed.
Out of the 10,889 women who took part in the trial, only 81 decided against having their scans reviewed by the AI system.
The AI tool, Mia, underwent training using a dataset from over 6,000 past breast cancer cases to identify the nuanced patterns and imaging markers indicative of malignant tumors. In its assessment of new cases, Mia achieved an 81.6 percent accuracy rate in predicting cancer presence and was 72.9 percent accurate in confirming its absence.
Breast cancer stands as the most prevalent cancer among women globally, with around two million new diagnoses each year. Although early detection and advancements in treatment have enhanced survival rates, a number of patients continue to endure severe side effects, such as lymphoedema following surgery and radiotherapy.
To address this, researchers are enhancing the AI system to forecast the likelihood of patients encountering such side effects up to three years post-treatment. This advancement could enable healthcare professionals to tailor care more effectively, offering alternative therapies or preventive measures for those identified as high-risk.
The research team is set to include 780 breast cancer patients in a clinical trial named Pre-Act, aiming to prospectively validate the AI’s risk prediction model through a two-year follow-up period. The overarching ambition is to develop an AI system capable of thoroughly assessing a patient’s prognosis and specific treatment requirements.
