AI Predicts Cancer Spreading To Brain from Lung Biopsy Images

In the study, scientists at Washington University School of Medicine in St. Louis (St. Louis, MO, USA) employed artificial intelligence (AI) to analyze lung biopsy images and predict the likelihood of the cancer spreading to the brain. Traditionally, pathologists have examined biopsy tissues under a microscope to spot signs of the disease. Now, AI seeks to emulate and enhance this diagnostic accuracy. The researchers trained a machine-learning algorithm with 118 lung biopsy samples from early-stage non-small cell lung cancer patients to predict brain metastasis. Some subjects later developed brain cancer over a five-year follow-up, while others went into remission.

Upon testing the AI on 40 additional patients, the researchers found that the algorithm impressively predicted brain cancer development with 87% accuracy, outperforming the average 57.3% accuracy among four pathologists involved in the study. The AI algorithm was particularly accurate in identifying patients who would remain free from brain metastasis. The algorithm evaluates tumors and healthy cells similar to how the brain recognizes familiar faces through facial features. Yet, the exact features the AI detects remain a mystery, prompting ongoing research to understand the molecular and cellular cues it uses to make predictions. This insight could revolutionize therapeutic development and guide the development of imaging tools tailored for AI data collection, potentially altering the treatment landscape for early-stage lung cancer patients.

“This study started as an attempt to find predictive biomarkers,” said Changhuei Yang, Ph.D., a professor of electrical engineering, bioengineering, and medical engineering at the California Institute of Technology. “But we couldn’t find any. Instead, we found that AI has the potential to make predictions about cancer progression using biopsy samples that are already being collected for diagnosis. If we can get to a prediction accuracy that will allow us to use this algorithm clinically and not have to resort to expensive biomarkers, we are talking about significant ramifications in cost-effectiveness.”

Related Links:
WashU Medicine