Researchers Developed An Artificial Intelligence (AI) Model That Successfully Predicts Cancer Patient Survival Rates By Reading Oncologist Notes
Researchers from the University of British Columbia and BC Cancer have successfully developed a novel artificial intelligence (AI) model that can predict the survival of cancer patients more accurately using more readily available data compared to previous techniques.
“Predicting cancer survival is an important factor that can be used to improve cancer care,” explained Dr. John-Jose Nunez, the study’s lead author.
“It might suggest health providers make an earlier referral to support services or offer a more aggressive treatment option upfront. Our hope is that a tool like this could be used to personalize and optimize the care a patient receives right away, giving them the best outcome possible.”
The model uses a branch of AI that comprehends complex human language– known as natural language processing (NLP)– to actually analyze notes taken by oncologists following patients’ initial consultations. This is typically the first step of the cancer battle after a diagnosis is issued.
The model was able to identify unique patient characteristics and predict survival rates at six months, 36 months, and 60 months with over 80 percent accuracy.
Cancer survival rates have traditionally been calculated retrospectively. They are also categorized based on a few general indicators– including tissue type and cancer site.
And in spite of oncologists’ familiarity with these survival rates, it can still be difficult to accurately predict a specific individual’s survival rate since there are various complex factors that may impact patient outcomes.
But, the AI model designed by the researchers is able to identify unique clues within patients’ initial consultation documents. This allows for a more refined assessment and is even applicable to all cancers– a major improvement compared to previous models, which have been limited to specific cancer types.
According to Dr. Nunez, the AI model reads consultation documents very similar to how humans would.
Svitlana – stock.adobe.com – illustrative purposes only, not the actual person
“These documents have many details like the age of the patient, the type of cancer, underlying health conditions, past substance abuse, and family histories. The AI brings all of this together to paint a more complex picture of patient outcomes,” he said.
In order to train and test the model, the researchers used data from 47,625 patients across six BC Cancer sites located in British Columbia. All patient data also remained securely stored at BC Cancer and was anonymously presented to protect privacy.
So, unlike chart evaluations conducted by humans, the AI model also offers the added benefit of complete patient confidentiality.
Now, the model was trained using British Columbia data– meaning that it is a robust tool for cancer survival predictions in that province. Moving forward, though, the researchers are confident that the technology could also be applied in other cancer clinics around the globe.
“The great thing about neural NLP models is that they are highly scalable, portable, and don’t require structured data sets,” Dr. Nunez detailed.
“We can quickly train these models using local data to improve performance in a new region. I would suspect that these models provide a good foundation anywhere in the world where patients are able to see an oncologist.”
To read the study’s complete findings, which have since been published in JAMA Network Open, visit the link here.
If true crime defines your free time, this is for you: join Chip Chick’s True Crime Tribe
Sign up for Chip Chick’s newsletter and get stories like this delivered to your inbox.
More About:Science