On January 1, 2020, Google Health teamed up with DeepMind, an artificial intelligence company, to launch an artificial intelligence breast cancer detection system in the leading academic journal Nature. The system has more capacity to detect breast cancer than a professional radiologist, or could help improve the accuracy and efficiency of breast cancer screening, the authors said. In order to detect breast cancer as early as possible, a number of countries have implemented large-scale breast screening programs. However, the accuracy of breast X-ray analysis by professional doctors still has high uncertainty, prone to misdiagnosis and missed diagnosis, resulting in diagnosis will cause anxiety and unnecessary invasive diagnosis.
For the study, Google’s technology director, Shravya Shetty, teamed up with researchers to train artificial intelligence deep learning models using two data sets. One data set contains 25,856 mammograms from the United Kingdom, while the other contains 3,097 mammograms from the United States. The results showed that the false positive rate of aia model test results was 5.7% (United States) and 1.2% lower than that of typical radiologists (United Kingdom), and the false negative rate was 9.4% (United States) and 2.7% lower than that of typical radiologists.
False positives, also known as misdiagnosis rates, refer to the percentage of people who are actually disease-free but are judged to be ill under screening. False negative rate, also known as the missed diagnosis rate, refers to the actual disease, but according to the screening test is classified as a disease-free percentage.
Artificial Intelligence Breast Cancer Screening System
Performance of artificial intelligence systems and clinicians in breast cancer prediction
Breast cancer is one of the most common cancers in women, affecting about 2.1 million women each year, according to the World Health Organization. In 2018, about 627,000 women worldwide died of breast cancer, about 15 percent of all cancer deaths.
Breast cancer prediction, artificial intelligence system compared to 6 independent filmmakers
In the independent sub-experiments in the study, artificial intelligence systems outperformed all six radiologists.
Breast screening in the UK is analysed by two radiologists. In response, the researchers found that using artificial intelligence systems reduced the workload of second-in-the-post doctors by 88 percent.
Google Health, DeepMind, University College London, Cambridge University, The Royal Surrey County Hospital in Guildford, Google start-up Verily Life Sciences, Stanford Medical Centre, Royal Marsden Hospital and others completed the study.
So far, artificial intelligence has tried to diagnose breast cancer.
In 2017, Google Medical AI outperformed professional pathologists in breast cancer diagnosis. The following year, Google released an artificial intelligence detection system for advanced breast cancer that correctly differentiates metastatic cancer in 99 percent of cases.
In October 2018, the Massachusetts Institute of Technology (MIT) released a deep learning model for the study of dense breast tissue in mammograms, similar to radiologists. In May 2019, MIT’s Computational Science and Artificial Intelligence Laboratory released an artificial intelligence system from Massachusetts General Hospital, which says it can predict whether a patient will develop breast cancer within the next five years from mammograms.
Breast cancer is not female-specific, and men also have a higher risk of breast cancer and higher mortality rates.
In September 2019, a large-scale study published in the Journal of the American Medical Association, the journal oncology, found that male breast cancer patients had a 19 percent higher mortality rate than women. The study noted that men with breast cancer were diagnosed at a much larger age than women, with an average age of 63.3 years and women at an average of 59.9 years. Male patients have a higher mortality rate at all stages than female patients. The researchers believe that clinical characteristics and undertreatment were associated with 63.3 percent of male mortality.