AI Technique Could Offer Tailored Cancer Treatments

When treating patients with early-stage lung cancer, physicians must decide whether to proceed with potentially beneficial but toxic therapies (like chemotherapy, radiation, or immunotherapy) to remove cancer and reduce the likelihood that it will spread to the brain or whether to wait and see if lung surgery alone is adequate.

When up to 70 % of these individuals do not develop brain metastasis (the spread of cancer to the brain), the question arises: who should receive further aggressive treatments, and who can wait?

For patients with early-stage lung cancer, new research from Washington University School of Medicine in St. Louis could assist physicians in finding the ideal mix between aggressive management and cautious surveillance. The study, published on March 4^th, 2024, in The Journal of Pathology, used artificial intelligence (AI) to analyze lung biopsy images from patients and predict the likelihood of brain metastases from cancer.

There are no predictive tools available to help physicians when treating patients with lung cancer. 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.

Richard J. Cote, MD, Edward Mallinckrodt Professor, Washington University School of Medicine in St. Louis

Lung cancer is the leading cause of cancer deaths in the United States and throughout the world. Most lung cancers are classified as non-small cell lung cancers, which are caused mainly by smoking. Tumors in early-stage cancer patients are limited to the lungs, and surgery is suggested as the first line of treatment.

Approximately 30 % of these individuals move to advanced stages, in which the disease has progressed to the lymph nodes and other organs. Since the brain is generally the first to be impacted, these individuals require extra treatments such as chemotherapy, targeted drug therapy, radiation therapy, and immunotherapy.

However, physicians have no way of predicting which patients’ cancers will grow, so they regularly treat them with severe therapies out of caution.

Cote collaborated with Ramaswamy Govindan, MD, the Anheuser Busch Endowed Chair in Medical Oncology and associate director of the oncology division at Washington University; Mark Watson, MD, Ph.D., the Margaret Gladys Smith Professor in the Department of Pathology & Immunology; and Changhuei Yang, Ph.D., a professor of electrical engineering, bioengineering, and medical engineering at the California Institute of Technology, to see if AI could predict whether cancer would spread.

Under a microscope, a pathologist analyzes biopsied tissues to look for cellular abnormalities that can indicate a disease during diagnostic testing. According to Cote, researchers are looking at advanced technologies like AI to mimic what pathologists perceive while making diagnoses more accurately.

Crucial question: Is it possible for AI to identify aberrant traits that a pathologist cannot?

Using lung biopsy data from 118 patients with early-stage non-small cell lung cancer, the researchers created a machine learning system to predict brain metastasis. During a five-year observation period, brain cancer manifested in some patients but not in others who were in remission.

The researchers next used lung biopsy samples from 40 more individuals to assess the AI method’s predictive power for brain metastasis and its capacity to identify patients who do not develop metastasis.

With 87 % accuracy, the algorithm was able to forecast when brain cancer will eventually manifest. In comparison, the average accuracy of the four pathologists who participated in the research was 57.3 %. Crucially, the system predicted which patients would not get brain metastases with high accuracy.

Our results need to be validated in a larger study, but we think AI has great potential to make accurate predictions and impact care decisions. 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.

Ramaswamy Govindan, MD, Anheuser Busch Endowed Chair in Medical Oncology, Washington University School of Medicine in St. Louis

The AI system analyzes the characteristics of healthy cells and tumors in a way akin to how the human brain scans facial features to identify well-known faces quickly. Scientists are attempting to comprehend the molecular and cellular features that AI employs for its predictions, but it is unclear what the algorithm sees.

This knowledge could contribute to developing new treatments and impact the creation of imaging devices that are best suited for collecting data for AI.

This study started as an attempt to find predictive biomarkers. 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.

Changhuei Yang, Ph.D., Professor, Washington University School of Medicine in St. Louis

The National Cancer Institute of the National Institutes of Health (NIH), grant numbers 5R01CA182746 and U01CA233363; Washington University in St. Louis School of Medicine Personalized Medicine Initiative; Sensing to Intelligence, grant number 13520296; and the Heritage Research Institute for the Advancement of Medicine and Science at Caltech, grant number HMRI-15-09-01 provided funding for the study.

Journal Reference:

Zhou, H., et. al. (2024) AI-guided histopathology predicts brain metastasis in lung cancer patients. The Journal of Pathology. doi:10.1002/path.6263.