Listen to this article AI Algorithms Predict Cancer Patient Survival Based On Analysis Of Doctor’s Notes
Introduction
Predicting cancer patient survival is crucial for personalized and optimized care. However, it can be challenging for oncologists to accurately predict individual patient outcomes due to the many complex factors that influence survival rates. In the past, experts calculated cancer survival rates by looking back at data and using only a limited set of generic factors, such as the type of cancer and where it originated in the body.

A team of researchers from the University of British Columbia and BC Cancer sought to address this challenge by developing an artificial intelligence algorithms AI algorithms predict cancer patient survival more accurately and with more readily available data than previous tools. Their findings were published in JAMA Network Open.
Natural Language Processing (NLP) is a branch of AI that understands complex human language. The team used NLP to analyze oncologist notes following a patient’s initial consultation visit – the first step in the cancer journey after diagnosis. The researchers demonstrated that the AI Algorithms Predict Cancer and can predict a patient’s survival rate accurately for different periods of time (six months, 36 months, and 60 months) with an accuracy of over 80%, by identifying unique characteristics of each patient.
Unique Characteristics Identified by the AI Model
The model developed by Dr. John-Jose Nunez and his collaborators, which includes researchers from BC Cancer and UBC’s departments of computer science and psychiatry, is able to pick up on unique clues within a patient’s initial consultation document to provide a more nuanced assessment. Previous models were only capable of predicting the survival rates for specific types of cancer, unlike this model which has the ability to predict the survival rates for all types of cancers.
The AI model essentially reads the consultation document similar to how a human would read it. Moreover, these documents have many details, such as the age of the patient, the type of cancer, underlying health conditions, past substance use, and family histories. Consequently, the AI brings all of this information together to paint a more complete picture of patient outcomes.
Training and Testing the Model
The researchers trained and tested the model using data from 47,625 patients across all six BC Cancer sites located in British Columbia. To ensure the privacy of patients, BC Cancer stored all patient data in a secure manner and presented it anonymously. Unlike chart reviews by human research assistants, the new AI approach has the added benefit of maintaining complete confidentiality of patient records.
Potential for Widespread Application
Dr. Nunez and his team’s technology has the potential to be applied in cancer clinics worldwide, including Canada. The model’s training on BC data makes it a powerful tool for predicting cancer survival in the province.
In the future, neural NLP models could be highly scalable, portable, and not require structured data sets. As a result, these models provide a good foundation anywhere in the world where patients are able to see an oncologist.
Improving Patient Care with AI
The BC Cancer Foundation provides funding to support Dr. Nunez, who is also the recipient of the 2022/23 UBC Institute of Mental Health Marshall Fellowship. In another stream of work, Dr. Nunez is examining how to facilitate the best-possible psychiatric and counselling care for cancer patients using advanced AI techniques. He envisions a future where AI is integrated into many aspects of the health system to improve patient care.
Moreover, the AI model developed by Dr. Nunez and his team has the potential to revolutionize cancer care. By providing a more accurate prediction of individual patient outcomes, AI acting as a virtual assistant for physicians will help sort through and make sense of all the data to inform physician decisions. This will ultimately improve the quality of life and outcomes for patients.
Comments 1