Researchers at NYU School of Medicine in the US have found that an AI program can analyse images of patients’ lung tumours and accurately specify cancer types and identify altered genes driving abnormal cell growth.

This is significant because it demonstrates that by using an AI approach to diagnosing the type of cancer patients will be able to start targeted cancer therapies sooner.

In a study published online in Nature Medicine, the team of researchers trained a deep convolutional neural network, Google’s Inception v3, to analyse slide images obtained from The Cancer Genome Atlas, a database of confirmed cancer diagnosis images. This enabled the researchers to measure how well their machine learning program could be trained to accurately and automatically classify normal versus diseased tissue.

From the study the researchers found that the AI program could distinguish with 97% accuracy between adenocarcinoma and squamous cell carcinoma, two lung cancer types that even experienced pathologists struggle to parse without confirmatory tests.

From analysing the images, the AI program was also able to determine if abnormal versions of six genes linked to lung cancer – including EGFR, KRAS and TP53 – were present in cells, with an accuracy that ranged from 73 to 86% depending on the gene. As the researchers explained in the study, these genetic mutations often cause the abnormal growth seen in cancer, but can also change a cell’s shape and interactions with its surroundings, providing visual clues for automated analysis.

For targeted therapies to be successful, it’s important to determine which genes are changed in each tumour. However, current genetic tests being used to confirm the presence of mutations can take weeks to return results.

“Delaying the start of cancer treatment is never good,” said senior study author and Associate Professor in the Department of Pathology at NYU School of Medicine and NYU Langone Health’s Perlmutter Cancer Center, Aristotelis Tsirigos, PhD.

“Our study provides strong evidence that an AI approach will be able to instantly determine cancer subtype and mutational profile to get patients started on targeted therapies sooner,” continued Tsirigos.

The study also found that about half of the small percentage of tumour images misclassified by the study AI program was also misclassified by the pathologists, demonstrating the difficulty in distinguishing between the two lung cancer types. Meanwhile, 45 out of 54 of the images misclassified by at least one of the pathologists in the study were assigned to the correct cancer type by the program, suggesting that AI could offer a useful second opinion.

“In our study, we were excited to improve on pathologist-level accuracies, and to show that AI can discover previously unknown patterns in the visible features of cancer cells and the tissues around them,” said co-corresponding author of the study and Assistant Professor in the departments of Radiology and Population Health, Narges Razavian, PhD.

“The synergy between data and computational power is creating unprecedented opportunities to improve both the practice and the science of medicine,” continued Razavian.

The research team plan to continue training their AI program with data until it can determine which genes are mutated in a given cancer with more than 90% accuracy. Following this they will seek US government approval to use the technology clinically.

For more information contact, like us on Facebook or tweet us @eHealthNewsZA.