One in eight women will receive a breast cancer diagnosis at some point in their lifetime. While it’s not the news anyone wants to hear, an accurate diagnosis is crucial in getting the right care. Traditionally, detection occurs after undergoing a mammogram screening and having a radiologist review your x-rays. But, sometimes the process can result in a false diagnosis or miss the cancer altogether. Despite some apprehension around the future of AI in healthcare, Google Health’s new algorithm is stepping in to try and solve the problem by improving the accuracy of detection.
“There are some promising signs that the model could potentially increase the accuracy and efficiency of screening programs, as well as reduce wait times and stress for patients,” said Google Health in a blog post.
To develop the AI, anonymized mammograms were collected from thousands of women in the US and UK. Researchers used the data to develop an algorithm that adapted to the results. The AI was then tested against the x-rays of different groups of women in both countries to analyze how often it was correct in determining whether a woman had breast cancer.
The results? Google’s AI outperformed a single radiologist in both the UK and US. In fact, it reduced false positives by 5.7 percent for women in the US and reduced false negatives by 9.4 percent – without the additional health data human radiologists have access to. While the results are nothing short of impressive, the AI is certainly not perfect. There were several cases of cancer that were caught by radiologists and not the AI.
"While this is exciting, early-stage research, validation in future trials is needed to better understand how models like these can be effectively integrated into clinical practice," said Dr. Mozziyar Etemadi, Northwestern study co-author, in a statement featured on CNET.
Why This Matters –
As we look to the future of AI in our industry, instances such as this one help to show the benefit of having tech work in tandem with health professionals. In cases of diagnosing serious health conditions such as cancer, AI’s role in easing the burden from radiologists while also providing increased accuracy and efficiency could make all the difference for patients and HCPs.
For example, in the UK the standard of care is to have two radiologists review a screening. While Google Health’s AI didn’t perform better than the radiologists combined, it did reduce the workload of the second reader by nearly 88 percent. At a time where there’s a shortage of radiologists, time-savings such as what was provided AI could make a drastic difference in the ability to see more patients, accurately diagnose and effectively treat patients earlier and decrease the need for a double screening review.
