AI Tool Developed at Northwestern University Offers Personalized Breast Cancer Treatment Plans

Breakthrough AI tool surpasses expert pathologists in predicting breast cancer outcomes, potentially reducing unnecessary chemotherapy

A team of researchers at Northwestern University Feinberg School of Medicine has introduced a groundbreaking artificial intelligence (AI) tool designed to enhance the precision of breast cancer treatment plans. The tool, developed using a deep learning approach, generates a Histomic Prognostic Signature (HiPS) risk score that outperforms evaluations conducted by expert pathologists, offering a more accurate prediction of individual patient outcomes.

The AI tool demonstrates the potential to identify breast cancer patients currently deemed high or intermediate risk who may become long-term survivors. By doing so, it opens the door to the possibility of reducing the duration or intensity of chemotherapy for these patients, thereby minimizing unpleasant side effects associated with the treatment.

In a Nature Medicine-published study titled “A population-level digital histologic biomarker for enhanced prognosis of invasive breast cancer,” the research team, led by Dr. Lee Cooper, Associate Professor of Pathology at Northwestern University Feinberg School of Medicine, highlighted the importance of considering non-cancerous elements in predicting patient outcomes. The AI model evaluates both cancerous and non-cancerous cells, showcasing the significance of non-cancer components that have often been overlooked in traditional evaluations.

Dr. Cooper emphasized, “Our study demonstrates the importance of non-cancer components in determining a patient’s outcome. The importance of these elements was known from biological studies, but this knowledge has not been effectively translated to clinical use.”

The AI system, analyzing 26 different properties of breast tissue, provides an overall prognostic score and individual scores for cancer, immune, and stromal cells. The interpretability and transparency of the HiPS score contribute to its effectiveness, allowing pathologists to understand the decision-making process of the AI model.

To train the AI model, the researchers enlisted the help of an international network of medical students and pathologists, creating a vast dataset of human-generated annotations from digital images of patient tissues. The study, conducted in collaboration with the American Cancer Society (ACS), utilized a unique dataset representing breast cancer patients from over 423 U.S. counties.

Dr. Cooper highlighted the potential impact of the AI tool, stating, “Adoption of the new model could provide patients diagnosed with breast cancers with a more accurate estimate of the risk associated with their disease, empowering them to make informed decisions about their clinical care.”

The next phase involves the prospective evaluation of the AI model for clinical use, coinciding with Northwestern Medicine’s transition to using digital images for diagnosis over the next three years. The researchers are also working on developing models for specific breast cancer types, further enhancing predictions and insights into the biology of breast cancers.