Artificial intelligence has been changing the way health care works. The AI algorithms are slowly adding a power boost to disease diagnosis and treatment. With a life-threatening disease like cancer, the major factor which can prove to be a life-saver is early detection and diagnosis. However, with conventional methods, early diagnosis has been a challenge. Now, artificial intelligence is making diagnosis easier and more accurate, thereby bringing a groundbreaking innovation.
Researchers at the National University of Singapore are using artificial intelligence technology for the identification of cancer cells. This is done by the acidity or pH levels of the cells using the AI algorithm. The technique can be of immense help in the detection of cancer cells in tissue samples obtained from blood tests like tumor biopsies or liquid biopsies.
This will be significant support for physicians since the number of cells in the samples can even exceed millions, thereby making the accuracy of the AI algorithm all the more highlighted.
How Does It Work
A pH-sensitive dye called bromothymol blue will be coated on the live cells. The dye will change color according to the levels of acidity. Each cell has its own unique ‘fingerprint,’ owing to the intracellular activity. This ‘fingerprint’ constitutes a combination of red, green, and blue components(RGB).
Cancer cells have lower acidic levels due to the altered pH. This leads to quite different RGB fingerprints as the cells react to the dye differently. A microscope with a digital color camera can then be used to capture these fingerprints. Then, the AI algorithm can be used to classify and image cells originating from one tissue into normal cells or cancer cells. The tests are time-effective with enhanced accuracy in comparison to the conventional testing methods. A test takes at most 35 minutes and the accuracy rate of the tests touches 95%.
According to Professor Lim Chwee Teck (Director, Institute for Health Innovation and Technology, NUS),
“This demonstrates the potential of our technique to be used as a fast, inexpensive and accurate tool for cancer diagnosis.”
The analysis was also used to draw lines of distinction between benign and metastatic cancer cells. Four different types of cell types were analyzed namely; normal cells, benign breast tumor cells, breast cancer cells, and pancreatic cancer cells. The identification and classification accounted for an accuracy rate of 93%.
The accuracy rates are highly dependent on the diversity of RGB fingerprints and on the number of cells that the algorithm is trained to identify.
A great plus of the technique is that it keeps the cells alive so that they could be used for drug tests in the future. This is in contrast to the current techniques of imaging which often lead to toxic effects on the cells, killing them off eventually.
The technique can prove to be quite effective and can be potentially extended to monitoring the progression of the disease, treatment, and also for keeping a check on potential relapse post-treatment.
