Stanford researchers look into AI tool to help fight cancer and other diseases
Researchers at Stanford University believe a newly developed AI tool could help in the fight against cancer and other diseases.
Olivier Gevaert, a biomedical data scientist at Stanford Medicine, showed CBS News Bay Area an example of a breast cancer tissue. Identifying and understanding the prevalence of specific genes within that tissue is just one hurdle in the fight against certain cancers.
"Specifically for breast cancer, gene sets and gene signatures are being used to determine treatment," said Gevaert.
Clinicians have increasingly guided the selection of which cancer treatments, including chemotherapies, immunotherapies and hormone-based therapies, to recommend to their patients based on not only which organ a patient's cancer affects, but also which genes a tumor is using to fuel its growth and spread.
Stanford Medicine's new AI program, SEQUOIA, can analyze an image from a tumor biopsy and rapidly determine what genes are likely turned on and off in the cells it contains.
"The red means that the gene is very active in that area, whereas the blue means that the gene is very inactive," said Gevaert.
Analyzing this information often requires costly and time-consuming genomic sequencing and testing.
"Sometimes molecular tests are being done, but that's very dependent on the cancer type and is rather uncommon," said Gevaert.
Experts said gene activity altering the appearance of cells is often imperceptible to a human eye.
So Gevaert and his team of researchers turned to artificial intelligence to find these patterns, and make cancer recurrence predictions.
"You can see the score is very low, so we are predicting, just from the image, that this patient is low risk," said Gevaert.
The researcher said there's "definitely room for improvement," but also believes it can help guide treatment plans and better predict patient outcomes.
The AI model can't yet be used in a clinical setting — it needs to be tested in clinical trials and be approved by the Food and Drug Administration before it's used in guiding treatment decisions.
For some cancer types, researchers said the AI-predicted gene activity had a more than 80% correlation with the real gene activity data.
"The performance of the model is tissue specific. So in certain cancers, it works very well, and in other cancers, there's definitely room for improvement," said Gevaert.
It's a tool that's being further refined, but even so, it was able to predict, in its infancy, the expression patterns of more than 15,000 different genes.
Gevaert said, "It allows to make predictions of 1000s of genes depending on the tissue for cancer patients from normally expensive molecular testing."
By identifying and mapping genes with the click of a button, Gecaert believes this tool could potentially speed up treatment decisions, and ultimately make those treatments more efficient.
Gevaert said his team is improving the algorithm and studying its potential applications.
He believes SEQUOIA could reduce the need for expensive gene expression tests.
The study was published in science journal Nature.
