|Using Artificial Intelligence, A Sensor Detects Cancer|
A Sensor sniffs out Cancer using Artificial Intelligence
Researchers at Memorial Sloan Kettering Cancer Center (MSK) have developed a sensor that can be trained to detect cancer, with the help of artificial intelligence.
Although the training does not work in the same way that a police dog is trained to sniff out explosives or drugs, the sensor does have some similarity to how the nose works. The nose can detect more than a trillion different odors, even though it only has a few hundred types of olfactory receptors. The pattern of which odor molecules bind to which receptors creates a kind of molecular signature that the brain uses to recognize a scent.
Like the nose, cancer detection technology uses an array of multiple sensors to detect a molecular signature of the disease. Instead of the signals going to the brain, they are interpreted by machine learning, a type of computer artificial intelligence.
MSK researchers led by Kravis WiSE postdoctoral fellow Mijin Kim and biomedical engineer Daniel Heller, head of MSK’s Cancer Nanomedicine Laboratory, built the technology using an array of sensors composed of carbon nanotubes. Carbon nanotubes are tiny tubes, nearly 100,000 times smaller than the width of a human hair. They are fluorescent, and the light they emit is very sensitive to tiny interactions with molecules in their environment.
Each nanotube sensor can detect many different molecules in a blood sample. By combining the many responses from the sensors, the technology creates a unique fluorescent pattern. The pattern can then be recognized by a machine learning algorithm that has been trained to identify the difference between a cancer fingerprint and a normal one.
In experiments conducted on blood samples obtained from ovarian cancer patients, the researchers found that their nanosensor detected ovarian cancer more accurately than currently available biomarker tests. (A biomarker is a particular chemical produced by tumors and spread through the bloodstream that indicates the presence of disease. In this case, the biomarker tests were for the ovarian cancer-related proteins CA125, HE4, and YKL40 ).
The hope for patients is that researchers will further develop the technology so that it can eventually be used in the clinic to rapidly detect early-stage ovarian cancer and many other types of cancer.
Need for better cancer screening tests
Tests that find cancers early using blood markers hold great promise for improving outcomes for people with cancer, especially those types, like ovarian cancer, that have few early signs or symptoms.
Several serum biomarker tests for ovarian cancer are already in use. Unfortunately, these independent biomarker measurements have proven ineffective in early detection. Currently, no screening strategy can identify ovarian cancer early enough to reduce mortality.
The nanosensor approach could potentially provide a better way.
“Ovarian cancer spreads along the surfaces of the abdomen and pelvis [rather than through the blood], which makes finding it with a blood test especially difficult,” says MSK surgeon Kara Long Roche, study author. “This technology could potentially find more subtle and complex changes in the blood, which may be the key to early detection, and early detection will save lives.”
To train the machine learning algorithm, the researchers needed to collect sensor responses from many blood samples, as this method requires many samples to be accurate. In addition, samples from patients with conditions other than ovarian cancer were used: “Certain other diseases can fool the sensor because they produce some of the same components in the blood,” says Dr. Kim, lead author of the study.
Although the technology improves the accuracy of ovarian cancer detection compared to current biomarker-based methods, more work is needed to enable early-stage ovarian cancer detection and confirm that this test works in people.
“We will not stop until there is a way to prevent ovarian cancer deaths,” says Dr. Heller.
Towards a Universal Cancer Sensor
The researchers say the technology can be adapted to detect many types of cancer using the same set of sensors, without the need to first identify a biomarker.
“A major limitation in the development of cancer screening tests has been the lack of sufficient biomarkers,” says Lakshmi Ramanathan, head of MSK’s Clinical Chemistry Service and an author of the study. “The ability to develop a screening test without the need for a biomarker is an exciting possibility for this type of technology.”
Furthermore, the same set of sensors could be used to train algorithms to detect different types of cancer. “We believe this technology can be developed to simultaneously detect many diseases, although additional measurements need to be made using samples from patients with these conditions,” says Dr. Heller.
Rekindle Cancer Moonshot
The technology comes as the Biden Administration restarts the Cancer Moonshot effort, now focusing it on cancer detection. Using a “universal” cancer sensor approach, the technology has the potential “to help make screening more accessible and available to all,” the report states.
Early detection, facilitated by cancer screening efforts, is a crucial strategy for preventing cancer deaths. As a recent White House report states: “With recommended regular screening tests, we can often find cancer when there may be more effective treatment options or even prevent it from developing by removing precancerous tissue.”
The article was published in the journal Nature Biomedical Engineering. In addition to those at MSK, the article has authors from Weill Cornell Medicine, Montefiore Medical Center, University of Maryland, Lehigh University, Hunter College High School, and the National Institute of Standards and Technology.
- Ovarian cancer is difficult to detect early because it causes few symptoms.
- Existing stand-alone biomarker tests for ovarian cancer are not effective in early detection.
- An AI-assisted nanosensor developed by MSK researchers can detect signs of ovarian cancer in the blood.
- Once validated, the test could potentially aid in the early detection of ovarian cancer.
This work was supported in part by NIH grants R01-CA215719, U54-CA137788, U54-CA132378, and P30-CA008748; the CAREER Award from the National Science Foundation (1752506); the Honorable Tina Brozman Foundation for Ovarian Cancer Research; Tina Brozman’s Ovarian Cancer Research Consortium 2.0; the Kelly Auletta Fund for Ovarian Cancer Research; the American Cancer Society Research Scholars Grant (GC230452); the Pershing Square Sohn Cancer Research Alliance; the Expect Miracles Foundation – Cancer Financial Services; the Center for Experimental Therapeutics; W. H. Goodwin and A. Goodwin and the Commonwealth Foundation for Cancer Research. Dr. Kim was supported by the Marie-Josée Kravis Women in Science Endeavor Postdoctoral Fellowship. Dr. Heller is a co-founder and officer, with equity participation, of Goldilocks Therapeutics Inc., Lime Therapeutics Inc., and Resident Diagnostics Inc., and is a member of the Scientific Advisory Board of Concarlo Holdings LLC, Nanorobotics Inc., and Mediphage Bioceuticals Inc. other authors declare no conflicting interests.