AI predicts oral cancer survival

The chances of surviving oral cancer can now be predicted from AI algorithms by measuring immune cells in tumors.

That’s according to the results of a pilot study conducted at the Department of Computer Science at the University of Warwick in England published in nature.

Researchers from the university developed a digital score to measure Tumor Infiltrating Lymphocytes (TILs), which can indicate chances of survival and also determine the stage of the cancer as well as predict the disease’s progression accurately. The more TILs present, the higher the chances of survival by showing a patient’s immunity to the cancer and response to treatment. Specifically, the density and spatial arrangement of TILs correlates to survival chances and disease-free survival.

The scans measuring the TILs were from images from patients at Shaukat Khanum Memorial Cancer Hospital Research Centre in Pakistan. They had already been treated with radiation and a head and neck surgery, and the cancer tissue samples were sent to University Hospital Coventry and Warwickshire. Researchers then used a state-of-the-art imaging machine to digital produce high-resolution images of the samples at microscopic scale.

“We are only beginning to unravel the remarkable potential of wealth of information present in pathology image data,” said Professor Nasir Rajpoot from the Department of Computer Science at the University of Warwick, who led the study. “This pilot study shows that with the help of modern cancer image analytics algorithms, we can precisely calculate the score of abundance of TILs in oral cancers in an objective manner and then use that score for risk stratification in terms of disease-free survival.”

Oral cancer is one of the most common types of head and neck cancers, particularly in certain parts of the world, such as South Asia, where is it more common to chew tobacco, consume betel quid or have higher rates of viral infections like human papillomavirus (HPV). The findings in the pilot study could help develop prognostication models to maker better treatment decisions.

This is a very exciting development. Not only is this one of the first artificial intelligence-based scores to be validated in oral cancer, this score also seems to have a strong prognostic power, which could eventually lead to stratifying patients for different treatment modalities,” said Hisham Mehanna, professor of Head & Neck Oncology at the University of Birmingham.