A study by Northwestern Medicine introduces a groundbreaking AI tool that enhances the prediction of breast cancer outcomes. This tool, which analyzes both cancerous and non-cancerous cells, may lead to more personalized and precise treatment strategies, potentially reducing the need for chemotherapy. The study encompasses a wide-ranging dataset and is oriented towards improving breast cancer diagnosis and treatment methods.
The AI tool developed by Northwestern Medicine is set to revolutionize breast cancer prognosis, with a focus on reducing unnecessary chemotherapy.
The tool has the potential to lessen disparities in diagnosis among patients in various community settings.
Non-cancerous cells are crucial in either supporting or hindering cancer growth.
Approximately one in eight women in the U.S. will be diagnosed with breast cancer during their lifetime.
A novel AI tool developed by Northwestern Medicine could significantly reduce the need for chemotherapy in breast cancer patients by offering a more accurate method of predicting patient outcomes.
The AI tool demonstrated superior capabilities in predicting the progression of a patient’s disease compared to traditional evaluations by expert pathologists.
This tool can identify breast cancer patients who, despite being categorized as high or intermediate risk, may have long-term survival prospects. This implies that the duration or intensity of their chemotherapy could be decreased, which is vital considering the severe side effects associated with chemotherapy, such as nausea or, in rare cases, cardiac damage.
Comprehensive AI Approach
Traditionally, pathologists focus solely on cancerous cells in a patient’s tissue for treatment determination. However, the study highlights the significance of non-cancerous cells in outcome prediction.
This research is the first to utilize AI in the comprehensive evaluation of both cancerous and non-cancerous components in invasive breast cancer.
“The significance of non-cancer elements in predicting patient outcomes is crucial, as shown by our study,” stated Lee Cooper, an associate professor of pathology at Northwestern University Feinberg School of Medicine. “Though these elements’ importance has been known from biological studies, their clinical application has been limited.”
The study will be published in Nature Medicine on November 27, 2023.
In 2023, about 300,000 U.S. women are expected to be diagnosed with invasive breast cancer.
The grading process, led by pathologists, assesses the abnormality of cancerous tissue. This traditional method, focusing on the appearance of cancer cells, has not seen significant changes for decades. The grade, as determined by the pathologist, informs the treatment plan for the patient.
Research in breast cancer biology indicates the vital role of non-cancerous cells, including immune system cells and structural cells, in either promoting or inhibiting cancer growth.
Cooper and his team developed an AI model that evaluates breast cancer tissue using digital images, analyzing both cancerous and non-cancerous cells and their interactions.
“These patterns are hard for pathologists to reliably categorize due to their complexity,” Cooper explained. “Our AI model quantifies these patterns and conveys the information in a way that is understandable for the pathologist.”
The AI system evaluates 26 different aspects of a patient’s breast tissue, resulting in a comprehensive prognostic score. It also provides individual scores for cancer, immune, and stromal cells, aiding pathologists in understanding the overall score. For instance, a favorable prognosis for some patients might be attributed to the properties of their immune cells, while for others, it could be due to their cancer cells. This information is vital for tailoring individualized treatment plans.
The adoption of this new model could provide patients with more accurate risk assessments, enabling them to make well-informed decisions regarding their treatment.
Moreover, this model could assist in monitoring therapeutic responses, potentially leading to adjustments in treatment intensity based on tissue changes over time. For example, it might identify the effectiveness of a patient’s immune response against cancer during chemotherapy, which could lead to a reduction in chemotherapy duration or intensity.
“We also aim to diminish disparities in diagnosis for patients in community settings,” added Cooper. “Our AI model could support generalist pathologists in evaluating breast cancers.”
Study Methodology
The study was a collaborative effort with the American Cancer Society (ACS), which provided a unique dataset of breast cancer patients from over 423 U.S. counties, primarily diagnosed or treated in community medical centers. Northwestern developed the AI software, while the ACS and National Cancer Institute contributed expertise in breast cancer epidemiology and clinical outcomes.
Training the AI model required extensive human-generated annotations of cells and tissue structures within digital patient tissue images. An international
