A large, rigorous study in Sweden of artificial intelligence (AI) in breast cancer screenings suggests AI can help doctors detect cancers more efficiently. We need more such studies to determine when the technology has real value — and when it might have risks.
Moreover, while the findings are incredibly promising, as Europe uses different processes and technologies for cancer screening, the US needs to commit to running its own similar studies to guide doctors here.
Previous large studies all looked back at old medical records to gauge whether AI could detect cancers as accurately as doctors. This study is the first trial of its size to test AI in real time on real patients — and would one day tell the field whether it actually improves the health of women. This is critical information as the technology increasingly becomes integrated into healthcare.
Illustration: Tania Chou
In the study, 80,000 women in Sweden were randomly assigned to either receive a double reading, where two independent radiologists look at the mammogram, or an AI-supported screening, which was performed by one radiologist and a computer.
The first stage of the study, the results of which were reported this week in Lancet Oncology, was designed to ask whether it was safe to integrate AI into practice. The answer is a resounding yes.
Overall, the computer helped humans flag more cancers, detecting about 20 percent more cancers than the two radiologists. Impressively, it did so with about the same rate as false positives (e.g., screens that looked like cancer, but did not turn out to be).
LOWER WORKLOAD
Moreover, the researchers clearly showed that AI can reduce the workload for radiologists. Although the team did not directly measure the number of hours saved by using a computer to analyze mammograms, they estimate that the technology reduced screen reading time by about 44 percent.
“In a situation where the medical workforce is strained, that’s a significant improvement,” said Larry Norton, medical director of the Evelyn H. Lauder Breast Center of Memorial Sloan Kettering Cancer Center.
Even if the technology does not turn out to be more accurate than doctors at picking up cancers, being just as accurate, but faster would still be a major advance, he said.
Now comes the hard work of showing that this improves cancer care.
“The holy grail is really understanding whether this kind of technology improves health,” said Yale School of Medicine’s Ilana Richman, whose research focuses on evaluating new breast cancer screening technologies. “We won’t know that for some time.”
The researchers in Sweden are to continue to study the women in their trial to try to answer that question. In addition to confirming AI’s performance in detecting cancers, they are set to probe whether those additional cancers being detected are meaningful — that is, are the additional early lesions caught by the computer ones that would eventually cause a woman harm?
They would also ask whether the method can reduce the number of “interval cancers,” or ones that are found between screenings and tend to be more aggressive and deadlier.
The need for this type of careful evaluation of AI is clear. So-called computer-aided detection that used more rudimentary versions of AI was widely adopted (particularly after the US Congress required Medicare to cover its use), but it led to an increase in false positives and biopsies for precancerous cells that are not typically dangerous.
All of that came at a cost to the healthcare system — when someone was flagged by a computer, they typically would go on to other types of tests and procedures that were not needed.
For now, any efficiencies that come out of the study would mostly benefit people in Europe and Australia, where breast cancer is typically screened by a team of two radiologists that might be safely reduced to one plus a computer.
Translating the results to the US is complicated by the different standard of care — mammograms are reviewed by just one radiologist and are typically a 3D scan rather than the 2D ones used in the Sweden study.
VALUABLE LESSONS
However, there are still some lessons for the US. For example, the algorithm used in the study was remarkably good at stratifying women by cancer risk — low, intermediate or high. And it turned out that the small number of women filtered into the high-risk group had a large portion of the cancers in the study.
That points to the potential to use AI to triage the sea of exams passing before a radiologist each day, helping them prioritize the high-risk ones. That could lead to patients getting treated as quickly as possible, said Laura Heacock, a radiologist at NYU Langone Health.
Patients might be wondering where all of this is leading: One day, will their cancer be diagnosed solely by a computer?
That is a prediction that Geoffrey Hinton, one of the so-called godfathers of AI, made back in 2016.
“If you work as a radiologist, you’re like the coyote that’s already over the edge of the cliff, but hasn’t yet looked down so doesn’t realize there’s no ground underneath him,” he said, suggesting they would be obsolete in five to 10 years.
“People should stop training radiologists now,” he said.
Seven years later, Hinton himself has warned of the potential dangers of AI and urged the field to proceed with more caution. Radiologists, meanwhile, have not gone anywhere. And their jobs should not be at risk — at least not until someone actually proves that AI is not just faster, but actually makes us healthier.
Until a study finds that in the US, technology would always be an addition to, rather than a replacement for, the deep expertise of a doctor.
Lisa Jarvis is a Bloomberg Opinion columnist covering biotech, healthcare and the pharmaceutical industry. Previously, she was executive editor of Chemical & Engineering News. This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Chinese Nationalist Party (KMT) caucus whip Fu Kun-chi (傅?萁) has caused havoc with his attempts to overturn the democratic and constitutional order in the legislature. If we look at this devolution from the context of a transition to democracy from authoritarianism in a culturally Chinese sense — that of zhonghua (中華) — then we are playing witness to a servile spirit from a millennia-old form of totalitarianism that is intent on damaging the nation’s hard-won democracy. This servile spirit is ingrained in Chinese culture. About a century ago, Chinese satirist and author Lu Xun (魯迅) saw through the servile nature of
In their New York Times bestseller How Democracies Die, Harvard political scientists Steven Levitsky and Daniel Ziblatt said that democracies today “may die at the hands not of generals but of elected leaders. Many government efforts to subvert democracy are ‘legal,’ in the sense that they are approved by the legislature or accepted by the courts. They may even be portrayed as efforts to improve democracy — making the judiciary more efficient, combating corruption, or cleaning up the electoral process.” Moreover, the two authors observe that those who denounce such legal threats to democracy are often “dismissed as exaggerating or
Monday was the 37th anniversary of former president Chiang Ching-kuo’s (蔣經國) death. Chiang — a son of former president Chiang Kai-shek (蔣介石), who had implemented party-state rule and martial law in Taiwan — has a complicated legacy. Whether one looks at his time in power in a positive or negative light depends very much on who they are, and what their relationship with the Chinese Nationalist Party (KMT) is. Although toward the end of his life Chiang Ching-kuo lifted martial law and steered Taiwan onto the path of democratization, these changes were forced upon him by internal and external pressures,
The Chinese Nationalist Party (KMT) caucus in the Legislative Yuan has made an internal decision to freeze NT$1.8 billion (US$54.7 million) of the indigenous submarine project’s NT$2 billion budget. This means that up to 90 percent of the budget cannot be utilized. It would only be accessible if the legislature agrees to lift the freeze sometime in the future. However, for Taiwan to construct its own submarines, it must rely on foreign support for several key pieces of equipment and technology. These foreign supporters would also be forced to endure significant pressure, infiltration and influence from Beijing. In other words,
RSS