Artificial intelligence that can spot early indicators of the disease years before clinicians would find it on a CT scan is on the verge of being developed, according to researchers in Boston, and will represent a significant improvement in lung cancer screening.
Scientists from the Massachusetts Institute of Technology and the Mass General Cancer Center in Cambridge created the new AI tool known as Sybil. A person’s likelihood of developing lung cancer in the coming year was demonstrated to be correctly predicted 86% to 94% of the time in one study.
Currently, the Centers for condition Control and Prevention advises adults who are at risk for lung cancer to undergo a yearly low-dose CT scan to screen for the condition.
The best radiologist’s eye can’t see everything, even with routine screening; this is where Sybil comes in.
Dr. Lecia Sequist, an oncologist and the director of the Cancer Early Detection and Diagnostics Clinic at Massachusetts General Hospital, asserted that the naked eye cannot see everything. The AI we created is looking at the scan in a totally different way than a human radiologist looks at it.
According to specialists, the technology could significantly improve the early diagnosis of lung cancer, which is the third most frequent cancer in the United States, per the CDC. The American Cancer Society predicts that this year will see more than 238,000 new instances of lung cancer and more than 127,000 deaths, making the condition the top cause of cancer-related death.
Sybil has not yet received FDA approval for usage outside of research studies, but if it does, it may have a special function.
Anant Madabhushi, a professor in the biomedical engineering department at the Emory University School of Medicine in Atlanta, claims that there are more than 300 artificial intelligence (AI) technologies that have been approved by the FDA for use in radiology. Most, he claimed, are employed to help medical professionals detect and treat cancer, but not to foretell a person’s future cancer risk, as Sybil did.
So that doctors know where to look and may find it as early as possible, Sybil searches for indications of where cancer is likely to appear.
According to Anant Madabhushi, a professor in the biomedical engineering division at Emory University School of Medicine in Atlanta, the FDA has given its blessing to more than 300 artificial intelligence (AI) radiology devices. Most, he asserted, are used to assist medical professionals in identifying and treating cancer, not to predict a person’s future chance of developing it, as Sybil did.
Sybil looks for signs of where cancer is likely to occur so that doctors know where to look and may discover it as early as possible.
Low-dose CT scan screening has been proven to cut the chance of dying from lung cancer by 24% because it can help find cancer earlier, when it is more curable.
However, Sandler noted that an AI tool might improve lung cancer early detection rates, which might also improve survival rates.
It’s excellent for radiologists, she said, but it’s even better for our patients to be able to use this tool to determine whether or not we think they have lung cancer.
How does Sybil operate?
Sybil uses a single CT scan to estimate the cancer’s risk. Dr. Florian Fintelmann, a radiologist at Mass General Cancer Center and one of the researchers working on Sybil, explained that it analyses the three-dimensional image in search of other patterns or annoyances that scientists are still trying to completely understand. These include indicators of abnormal growth in the lungs.
Sybil can predict a person’s likelihood of developing lung cancer in the next one to six years based on what it sees.
Fintelmann continued that there have been instances where Sybil has picked up cancer signals that radiologists had missed until nodules appeared on a CT scan years later.
In the future, according to Fintelmann, the AI technology will assist radiologists in making crucial treatment decisions rather than completely replace them.
He declared that artificial intelligence will assist radiology in the future. You will still need a radiologist to locate the cancer, choose the best course of action, and carry out the course of action.
But AI is still a long way from being perfect.
The type of data being used to train AI is one issue that Madabhushi, of Emory University, says he is concerned about.
A lot of data obtained from medical institutions or clinical trials do not represent the diversity of our country, he asserted, adding that, in his opinion, the AI tools are not being developed in a way that is specifically designed to assist Black and brown people.
The researchers who created Sybil stated that there are not enough Black or Hispanic patients in the data to have confidence in broad applicability yet.
By declaring last year that it would soon require researchers and businesses seeking approval for medical devices to submit a plan that ensures diversity in clinical trials, the FDA has already taken a move to address this issue.
We need to make sure that AI doesn’t reflect or pick up on our prejudices, Madabhushi continued.
Overdiagnosis is another issue, according to Sandler. For a nodule that may turn out to be benign, doctors may subject patients to a possibly unneeded biopsy.
Are we potentially discovering cancer when we don’t need to?
