The name Sybil originates from the oracles of ancient Greece, also known as sibyls: female figures trusted to convey divine knowledge of the hidden and omnipotent past, present, and future. Now, the name has been dug up since ancient times and given to an artificial intelligence tool for lung cancer risk assessment being developed by researchers at MIT’s Abdul Latif Jameel Clinic for Machine Learning in Health, the Mass. General Cancer Center (MGCC) and Chang Gung Memorial Hospital. (CGMH).
Lung cancer is the number 1 deadliest cancer in the world, resulting in 1.7 million deaths worldwide in 2020killing more people than the next three deadliest cancers combined.
“It’s the biggest killer of cancer because it’s relatively common and relatively difficult to treat, especially once it’s reached an advanced stage,” says Florian Fintelmann, MGCC interventional thoracic radiologist and co-author of the new paper. “In this case, it’s important to know that if you catch lung cancer early, the long-term outcome is significantly better. Your five-year survival rate is closer to 70 percent, whereas if you catch it when it’s advanced, the five-year survival rate is just under 10 percent.”
Although there has been an increase in new therapies introduced to combat lung cancer in recent years, the majority of lung cancer patients still succumb to the disease. Low-dose computed tomography (LDCT) scans of the lung are currently the most common way patients are screened for lung cancer in the hope of finding it in the early stages, when it can still be surgically removed. . Sybil takes the assessment a step further, analyzing LDCT imaging data without the assistance of a radiologist to predict a patient’s risk of developing future lung cancer within six years.
in his new article published in the Journal of Clinical Oncology, Jameel Clinic, MGCC, and CGMH investigators demonstrated that Sybil achieved C-scores of 0.75, 0.81, and 0.80 over the course of six years from various sets of lung LDCT scans taken from the National Lung Cancer Screening Trial (NLST), Mass General Hospital (MGH), and CGMH, respectively: Models that achieve a C-Index score greater than 0.7 are considered good, and greater than 0.8 are considered strong. The ROC-AUC for the one-year prediction using Sybil scored even higher, with a range of 0.86 to 0.94, with 1.00 being the highest possible score.
Despite its success, the 3D nature of lung CT scans made Sybil a challenge to build. Co-author Peter Mikhael, an MIT electrical and computer engineering doctoral student affiliated with the Jameel Clinic and the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), compared the process to “trying to find a needle in a haystack”. The imaging data used to train Sybil was largely free of signs of cancer because early-stage lung cancer occupies small portions of the lung, just a fraction of the hundreds of thousands of pixels that make up each CT scan. The densest portions of lung tissue are known as lung nodules and while they have the potential to be cancerous, most are not and can occur from healed infections or airborne irritants.
To ensure that Sybil could accurately assess cancer risk, Fintelmann and her team labeled hundreds of CT scans with visible cancerous tumors that would be used to train Sybil before testing the model on CT scans with no discernible signs of cancer.
Jeremy Wohlwend, a doctoral student in electrical engineering and computer science at MIT, a co-author of the paper, and an affiliate of the Jameel Clinic and CSAIL, was surprised by Sybil’s high score despite the lack of visible cancer. “We discovered that while [as humans] couldn’t see exactly where the cancer was, the model might still have some predictive power about which lung would develop cancer over time,” he recalls. “Knowledge [Sybil] he was able to highlight which side was the most likely, it was really interesting for us.”
Co-author Lecia V. Sequist, a medical oncologist, lung cancer expert, and director of the Center for Cancer Early Detection Innovation at MGH, says the results the team achieved with Sybil are important “because screening for lung cancer lung is not implemented to its full potential in the US or globally, and Sybil can help us close this gap.”
Lung cancer screening programs are underdeveloped in the regions of the United States most affected by lung cancer due to a variety of factors. These range from stigma against smokers to political and regulatory landscape factors, such as Medicaid expansion, which vary from state to state.
In addition, many patients diagnosed with lung cancer today have never smoked or are former smokers who quit smoking more than 15 years ago, traits that make both groups ineligible for lung cancer CT screening in recent years. USA.
“Our training data consisted of smokers only because this was a necessary criteria for enrollment in the NLST,” says Mikhael. “In Taiwan, they test nonsmokers, so our validation data is expected to include nonsmokers, and it was exciting to see that Sybil generalized well to that population.”
“An exciting next step in research will be to prospectively test Sybil in people at risk for lung cancer who have not smoked or who quit smoking decades ago,” says Sequist. “I treat these patients every day in my lung cancer clinic and it is understandable that they find it difficult to accept that they would not have been candidates for screening. Maybe that will change in the future.”
There is a growing population of lung cancer patients who are classified as non-smokers. Women who do not smoke are more likely to be diagnosed with lung cancer than men who do not smoke. Worldwide, more than 50 percent of women diagnosed with lung cancer do not smoke, compared to 15 to 20 percent of men.
MIT professor Regina Barzilay, co-author of the paper and Jameel Clinic AI faculty leader, who is also a fellow at the Koch Institute for Integrative Cancer Research, credits the joint MIT and MGH efforts on Sybil to Sylvia, the sister of a close friend of Barzilay’s and one of Sequist’s patients. “Sylvia was young, healthy and athletic; she never smoked,” recalls Barzilay. “When she started coughing, neither her doctors nor her family initially suspected that the cause might be lung cancer. When Sylvia was finally diagnosed and met Dr. Sequist, the disease was too advanced to reverse its course, when we mourned her death. from Sylvia, we couldn’t stop thinking about how many other patients have similar trajectories.”
This work was supported by the Bridge Project, a partnership between the Koch Institute at MIT and the Dana-Farber/Harvard Cancer Center; the MIT Jameel Clinic; quantum computer; Facing cancer; the MGH Center for Innovation in Cancer Early Detection; the Bralower and Landry families; Higher stage lung cancer; and the Eric and Wendy Schmidt Center of the Broad Institute of MIT and Harvard. The Linkou CGMH Cancer Center under the Chang Gung Medical Foundation provided assistance with data collection and R. Yang, J. Song and their team (Quanta Computer Inc.) provided technical and computer support to analyze the CGMH data set. . The authors thank the National Cancer Institute for accessing the NCI data collected by the National Lung Screening Trial, as well as the patients who participated in the trial.
