We are searching data for your request:
Upon completion, a link will appear to access the found materials.
Artificial intelligence to diagnose breast cancer?
Unfortunately, many women around the world still develop breast cancer every year, which is discovered relatively late. Improved diagnosis is needed. Artificial intelligence from Google is now able to detect breast cancer more precisely on mammography than radiologists.
The recent joint study by Google Health, Northwestern University and the Royal Surrey County Hospital found that artificial intelligence can identify breast cancer more accurately than radiologists. The results of the study were published in the English-language journal "Nature".
Early detection and diagnosis of breast cancer is difficult
Breast cancer is the most common cancer in women. Digital mammography or chest X-rays are used here for early detection. Despite the widespread use of mammography, the early detection and diagnosis of breast cancer remains problematic.
Evaluation of X-ray images can lead to incorrect results
Correct evaluation of the X-ray images is a difficult task even for people trained on them, which can lead to both false positive and false negative results. Inaccuracies can lead to delays in diagnosis and treatment and to unnecessary stress for the women concerned. In addition, they lead to a higher workload for radiologists who are already fully occupied.
Can artificial intelligence improve the diagnosis of breast cancer?
Over the past two years, collaborations with leading clinical research partners in the UK and US have reviewed whether artificial intelligence could improve breast cancer detection.
AI was more reliable than radiologists
The results presented show that the artificial intelligence model in so-called de-identified screening mammographies (in which identifiable information was removed) recognizes breast cancer with greater accuracy, less false-positive and less false-negative results than radiologists.
How was the AI trained and coordinated?
This creates the conditions for future applications in which the model could possibly assist radiologists in performing breast cancer screenings. The new model was trained and matched using a representative dataset consisting of de-identified mammographies from more than 76,000 women in the UK and more than 15,000 women in the US.
System achieved reduction in false results
The model was then tested on a separate de-identified data set of more than 25,000 women in the UK and over 3,000 women in the US. In this evaluation, the system achieved a false positive reduction of 5.7 percent in the US and 1.2 percent in the UK.
Another experiment confirms results
It was also checked whether the model was transferable to other health systems. In this separate experiment, there was a 3.5 percent reduction in false positives and an 8.1 percent reduction in false negatives, the researchers report.
AI needs less information for diagnosis
It is also noteworthy that the model requires less information to make its decisions. Human radiologists had access to patient history and previous mammographies, while the model only processed the most recent anonymized mammogram without additional information. Despite this limitation, the model outperformed individual radiologists in accurately identifying breast cancer.
More research is needed
Looking ahead, there are some promising signs that the model could potentially increase the accuracy and efficiency of screening programs and reduce waiting times and stress for women. However, further research, prospective clinical trials and regulatory approval are required to achieve this. (as)
Author and source information
This text corresponds to the specifications of the medical literature, medical guidelines and current studies and has been checked by medical doctors.
Swell:
- Scott Mayer McKinney, Marcin Sieniek, Varun Godbole, Jonathan Godwin, Shravya Shetty et al .: International evaluation of an AI system for breast cancer screening, in Nature (query: 06.01.2020), Nature
