Let There Be Light
Researchers develop artificial intelligence-aided technology for early cancer detection using lightBy Gayatri Namboodiri
Any news in cancer research is good news as it means one step forward in the direction of curing the disease. Recently, researchers from the Indian Institute of Science Education and Research (IISER), Kolkata, and the Indian Institute of Technology (IIT), Kanpur, developed a tool using artificial intelligence, which can detect cancer with the help of light. This technology can detect cancer in its early stages and that too, within a span of few minutes. This path-breaking research will prove to be very helpful in the early detection of cancer, thus saving many precious lives.
Explaining the basis behind the technology, Sabyasachi Mukhopadhyay from IISER Kolkata explained that cells that are progressing towards cancer show more complex geometric patterns than the normal, unaffected, healthy ones.
“It is on the basis of this correlation that this novel light-scattering based method was designed to identify these unique patterns for detecting cancer progression,” he said. Mukhopadhyay is the first author of a paper detailing this research development, which was published in the Journal of Biomedical Optics.
How the Technology Works
Giving more details about how the AI technology works, Mukhopadhyay said that white light spectroscopy (340nm-800nm) has been used in this research. At the different classification stages of cancer like binary or multi class classification, machine learning algorithm like Hidden Markov Model (HMM) and Support Vector Machine (SVM) are used.They have an important role in optimum accuracy production on the basis of strength of training data-sets (labelled database).”
By using a light-based probe to identify certain complex repeating patterns (called multi-fractals), present on developing cancer cells, we can get early indications of the disease,” said Mukhopadhyay. Hence, state-of-the-art signal processing tools like multi-fractal de-trended fluctuation analysis (MFDFA) have played the role of a statistical bio-marker in order to extract the multi-fractal parameters like Hurst exponent and width of singularity spectrum, he added.
In the research, the machine learning algorithms like Hidden Markov Model and Support Vector Machine are applied on multifractal parameters in order to classify healthy and abnormal tissues. In a normal tissue, the microstructure is uniform but as the cancer develops, this microstructure changes and it is on the basis of this correlation that the light scattering method identifies these microstructures to detect cancer. Another interesting observation is the fact that the MFDFA-HMM integrated model functions in a better manner than MFDFA-SVM model.
The research team has tested these algorithms in vivo and vitro samples. In vivo sample studies are those done on living animals, humans and plants. In vitro translates into “in glass”. Both the studies are done on samples outside their biological contexts. The accuracy of the algorithms was found to over 90% and 95% in vivo and in vitro samples respectively. There are plans to expand these tests and conduct further investigations.
“We are expanding our ‘in vivo’ study in order to deploy the technology for clinical trials in hospitals,” says Mukhopadhyay. Talking about the major challenges faced during the research,Mukhopadhyay said that the major challenge was to have adequate datasets to train by machine learning tools like SVM, HMM. The team is very hopeful that this research will contribute a lot to early cancer detection and treatment. The other researchers involved in this study included Dr Nandan Kumar Das (Postdoctoral Research Fellow at National University of Ireland, Galway), Dr Nirmalya Ghosh and Prof. Prasanta K Panigrahi from IISER Kolkata as well as Indrajit Kurmi and Prof. Asima Pradhan of IIT Kanpur.
Artificial Intelligence, which can detect cancer cells in its early stages using light within very less time. Researchers have developed algorithms that can track the cancer cells, which are not visible to the naked eye. These algorithms can also identify the stage of cancer, with 95% accuracy
The research was jointly done by the Indian Institute of Science Education and Research, (IISER), Kolkata and the Indian Institute of Technology, (IIT) Kanpur.
Sabyasachi Mukhopadhyay, IISER Kolkata Dr Nandan Kumar Das, Postdoctoral Research Fellow, National University of Ireland, Galway Dr Nirmalya Ghosh, IISER Kolkata Prof. Prasanta K Panigrahi, IISER Kolkata Indrajit Kurmi, IIT Kanpur Prof. Asima Pradhan, IIT Kanpur
Researchers hope to use the ‘in vivo’ sample studies for clinical trials in hospitals. Plans are also on to increase the scope of the tests further.