Deep learning tool predicts brain metastasis in lung cancer patients

A research team from Washington University School of Medicine in St. Louis has developed a deep learning (DL)-based approach to help predict which patients with non-small cell lung cancer (NSCLC) are likely to experience brain metastasis, according to research published last week in The Journal of Pathology.

The researchers emphasized that brain metastases can occur in almost half of patients with early and locally advanced NSCLC, but there are currently no reliable means to flag these patients before the cancer spreads to the brain.

For those with early-stage NSCLC, surgery is typically recommended as the first line of treatment, and additional therapies are considered once the cancer has spread to other organs and the lymph nodes.

Brain metastasis requires aggressive treatments like radiation therapy, immunotherapy, chemotherapy and targeted drug therapy. However, since clinicians have no way of knowing whether a patient’s cancer will spread to the brain, these therapies are often utilized as precautionary measures.

This approach can expose patients to treatments that they may not need, potentially leading to adverse outcomes.

To address this, the research team set out to develop an artificial intelligence (AI)-driven model to predict brain metastasis risk using lung biopsy images.

“There are no predictive tools available to help physicians when treating patients with lung cancer,” said Richard J. Cote, MD, the Edward Mallinckrodt Professor and head of the Department of Pathology & Immunology at Washington University, in a news release. “We have risk predictors that tell us which population is more likely to progress to more advanced stages, but we lack the ability to predict individual patient outcomes. Our study is an indication that AI methods may be able to make meaningful predictions that are specific and sensitive enough to impact patient management.”

The study aimed to shed light on whether an AI tool could detect abnormal features within a biopsy image that a pathologist might miss.

The algorithm was trained to predict brain metastasis using 118 lung biopsy samples from early-stage NSCLC patients who either did or did not develop brain cancer during a five-year monitoring period. The model was tested using an additional set of lung biopsy images from 40 other patients.

The algorithm was compared to four expert pathologists.

The analysis revealed that the DL model significantly outperformed the clinicians, achieving 87 percent accuracy in predicting brain metastasis compared to the pathologists’ average of 57.3 percent. The model was particularly accurate in flagging which NSCLC patients would not go on to develop brain cancer.

“Our results need to be validated in a larger study, but we think there is great potential for AI to make accurate predictions and impact care decisions,” said Ramaswamy Govindan, MD, the Anheuser Busch Endowed Chair in Medical Oncology and associate director of the oncology division at Washington University. “Systemic treatments such as chemotherapy, while effective in killing cancer cells, can also harm healthy cells and are not always the preferred treatment method for all early-stage cancer patients. Identification of patients who are likely to relapse in the brain may help us develop strategies to intercept cancer early in the process of metastasis. We think AI-based predictions could, one day, inform personalized treatments.”

The features that the tool uses to guide its predictions are unknown, but the researchers plan to explore this in future studies.

“This study started as an attempt to find predictive biomarkers,” explained Changhuei Yang, PhD, a professor of electrical engineering, bioengineering and medical engineering at the California Institute of Technology, who contributed to the study. “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.”

The use of AI in cancer research is rising as the technology continues to advance.

This week, Mayo Clinic researchers unveiled a new class of AI to help bolster oncology research.

Their ‘hypothesis-driven AI’ is designed to enhance knowledge discovery for diseases like cancer by allowing research teams to incorporate a specific hypothesis or research question into the algorithm. In this way, these models could leverage existing scientific knowledge, rather than relying strictly on static, difficult-to-obtain datasets.

The researchers posited that the approach holds significant promise to surface insights missed by traditional AI models.