Google has been working with clinical research partners in the United Kingdom and the United States to see if artificial intelligence (AI) could be used to improve the detection of breast cancer.
In collaboration with the Cancer Research UK Imperial Centre, Northwestern University, and Royal Surrey County Hospital, Google has created an AI model for reading mammograms, which are X-rays of the breast, to help radiologists spot the signs of breast cancer more accurately.
According to the American Cancer Society, mammograms miss about 20% of breast cancers in the United States, and false positives are common, resulting in women being called back for more tests, sometimes even biopsies.
"Reading these X-ray images is a difficult task, even for experts, and can often result in both false positives and false negatives. In turn, these inaccuracies can lead to delays in detection and treatment, unnecessary stress for patients and a higher workload for radiologists who are already in short supply," Google said.
Google's AI model was trained on a representative dataset comprised of de-identified mammograms from more than 76,000 women in the United Kingdom and over 15,000 women from the United States to see if it could learn to spot signs of breast cancer in the scans.
To determine its accuracy, the model was tested with a separate set of mammograms where the diagnosis was already known. Its results were then compared to the performances of the radiologists who had originally read the X-rays.
According to Google, the model's results were better than radiologists from both countries. From reading scans of 3,000 women in the United States, the model produced a 5.7% reduction of false positives and a 9.4% reduction in false negatives. On 25,000 mammograms performed in the United Kingdom, the system reduced false negatives by 2.7% and false positives by 1.2%.
It was also able to achieve this, Google said, despite having less information than human experts did when making its decisions. The human experts had access to patient histories and prior mammograms, while the model only processed the most recent anonymised mammogram with no extra information.
Google also performed a separate test to see if the model could make readings from generalised mammogram data. In this separate test, Google trained the model only on data from the women in the United Kingdom, and then evaluated it on the dataset from women in the United States. This resulted in a 3.5% reduction in false positives and an 8.1% reduction in false negatives, showing that the model still performed at a higher level than experts despite using only generalised mammogram data.
"Looking forward to future applications, 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," Google said.
Google has not been alone in developing AI projects for cancer detection. IBM in July released three AI projects tailored to the challenge of curing cancer to the open-source community.
Meanwhile in Australia, analytics company Max Kelsen is using AI to predict the effectiveness of cancer treatments. The company is integrating AI and whole-genome sequencing into cancer research and clinical practice, focusing initially on immunotherapy treatment for melanoma and small cell lung cancer.
The Melbourne-based centre used artificial intelligence to match lung cancer patients to relevant clinical trials.
If you love open source so much, say sorry for the horrible things you said, software pioneer Stallman suggests.
Partnerships between public health services and big tech will bring new treatments to patients faster. But sharing data is still the big stumbling block.
One way to speed up research? Steal it.
Why half of hospitals plan to invest in AI by 2021 (TechRepublic)
Artificial intelligence and RPA investments are slowly growing in the healthcare sector, according to a report from Olive.