Singapore researchers using AI to identify cancer cells by acidity
By Miro Lu
Although cancer cells and normal cells have many key differences, they can look rather similar under a microscope.
A research team from the National University of Singapore (NUS) has developed a technique that uses artificial intelligence to determine whether a single cell is healthy or cancerous by analyzing its pH level.
The research, led by Professor Lim Chwee Teck, director of the Institute for Health Innovation & Technology (iHealthtech) at NUS, was first published in the journal APL Bioengineering on March 16.
"We know that cancer cells are typically less acidic compared to other healthy cells, so we developed a technique where we stain the cell with pH-sensitive dye, and thereafter you use a simple microscope and digital color camera to capture the image. To identify whether they are cancer cells or not, we actually developed a machine algorithm to look at the unique acidic signature that is displayed in a cancer cell versus non-cancerous cells," said Lim, who is also from the Department of Biomedical Engineering at NUS.
The research team comprises Lim and two other research fellows, Dr. Yuri Belotti and Dr. Jokhun Doorgesh Sharma. They explained to CGTN that the current technique of identifying cancer cells requires tumor biopsy, then clinicians need to first stain the cells with fluorescent to see if the cells light up. The process takes a few hours to a day or two. It also requires more sophisticated equipment.
The NUS team says their cancer test can be completed in under 35 minutes, and single cells can be classified with an accuracy rate of more than 95 percent.
They envision a real-time technique where clinicians are able to diagnose cancer at any stage based on the sample obtained from a blood test.
"In terms of application, what we hope is that first from the patient we can perform liquid biopsies. After that, if we can isolate cancer cells from the blood cells, then we can subsequently perform more tests on these cancer cells," said Lim.
Isolating cancer cells from blood cells is no easy feat.
In 1 milliliter of blood, there are 5 billion red blood cells and just 10 to 100 cancer cells. For years researchers have been peering down microscopes, looking for distinct features that can help them determine the differences between a cancerous cell and a normal cell. "But naked eyes can fail us, whether the person is young or old, sometimes you see things quite differently, especially subtle differences in terms of the features."
The researchers believe that machine learning is playing an important role in their field of work.
The research team hopes this quick and low-cost technique can be extended to monitor cancer progression, effectiveness of a treatment, and even alerting the risk of a relapse after a successful treatment.