Google AI Beats Doctors On Breast Cancer Screening Accuracy

Google AI Beats Doctors On Breast Cancer Screening Accuracy

The accuracy of breast cancer screening is set to make a significant leap forward after a Google Health AI system demonstrated better results than those achieved by human specialists. The company has published a research paper in the scientific journal Nature, presenting the results of a joint study conducted by experts from Google Health, Google-parent’s AI unit DeepMind and scientists from leading UK and US universities.

The Google AI model was able to generate both fewer false positives and false negatives when assessing screening mammograms. Current methods that rely on human specialists analysing the results of mammograms lead to the failure to detect one in five incidents of breast cancer. Almost half of all women also receive an average of one false positive result every ten years, causing them significant stress in worry. Unnecessary treatment resulting from false positive results is estimated to cost more than $4 billion a year in the USA alone.

Google Health’s research was authored together with the Cancer Research UK Imperial Centre, Northwestern University in Illinois, and the Royal Surrey County Hospital. The AI model reduced the number of false positives by 5.7 per cent in the US and 1.2 per cent in the UK. It also reduced false negatives by 9.4 per cent in the US, and 2.7 per cent in the UK.

Dominic King, the UK lead for the Google Health unit described the research’s results as “really exciting”.

The DeepMind-developed screening algorithm was trained on almost 120,000 mammograms from patients in both the UK and USA. Google Health took over the project following the recent transfer of DeepMind’s health unit to Google. Dr King, whose previous career was as a breast cancer surgeon explained that the study was incentivised by worries current cancer screening services are unsustainable. The UK’s Royal College of Radiologists estimated in 2018 that maintaining current breast cancer screening systems would require over 1000 more full-time diagnostic radiologists than are currently available.

The hope is that AI screening will eventually be able to help take over much of the work needed to provide effective and improved cancer screening services. However, the important step of regulatory approval will first have to be secured. He commented:

“It could be a second opinion, giving a nudge or a recommendation here and there to spend more time looking at this scan, or to flag examples where cases get missed.”

AI is demonstrating increasing promise in the sphere of healthcare. Spotting patterns on the kind of images used in fields including pathology, ophthalmology and dermatology is considered as particularly well suited to the competencies of AI algorithms that use computer vision.

As well as the breast cancer screening algorithm under development, Google has already created the Lymph Node Assistant. It has demonstrated 99% accuracy in detecting late-stage breast cancer cells that have spread around the body. DeepMind will also soon reach the commercial launch of a device that can diagnose complex eye diseases as well as specialists.

Big tech are not the only companies developing AI for health applications. Start-up Kheiron Medical is also working on a breast cancer detection algorithm that has performed well in early trials.

Leave a Comment