A research team at Stony Brook University is working on a new approach to breast cancer diagnosis and treatment. The group, led by Chao Chen, PhD, and Prateek Prasanna, PhD, from the Department of Biomedical Informatics in the Renaissance School of Medicine and College of Engineering and Applied Sciences, aims to improve imaging analysis using mathematical modeling and deep learning.
The researchers focus on understanding breast tissue architecture changes over time. They note that while high breast density is a known risk factor for breast cancer, the complexity of breast tissue can make it difficult for clinicians to detect subtle changes using standard imaging methods.
To address these challenges, Chen and Prasanna are developing "TopoQuant," a suite of informatics tools designed to analyze breast tissue images. TopoQuant uses advanced mathematical modeling and machine learning to study the structural complexity of breast parenchyma. The goal is to collaborate with Stony Brook Medicine clinicians to better understand how tissue architecture changes during cancer development, progression, and treatment.
This project is supported by a four-year $1.2 million grant from the National Cancer Institute (NCI), running through August 2028. Both Chen and Prasanna are affiliated with the Imaging, Biomarker Discovery, and Engineering Sciences Research Division at the Stony Brook Cancer Center.
Chen explains that "this research will offer new insights into how structural changes in breast tissue can influence cancer screening and treatment outcomes." He highlights the use of topology in combination with deep learning as a means to develop novel algorithms that could lead to improved predictive models and treatment strategies.
Existing machine learning-driven tools used in cancer imaging lack interpretability or explanation capacity according to Stony Brook investigators. However, they assert that TopoQuant will provide clinicians with quantitative evidence regarding changes in breast tissue architecture related to cancer risk and treatment response.
In 2021 preliminary findings showed the efficacy of their approach using one informatics tool in predicting patient responses to neoadjuvant chemotherapy for breast cancer. The results indicated differences in topological behavior between patients who responded well to therapy versus those who did not.
Prasanna notes that their prediction models are unique because they ensure interpretability by design rather than relying on traditional post-hoc interpretation. He adds that while this research focuses on improving breast cancer diagnosis and treatment, it may also have applications in fields like neuroscience due to potential cross-disciplinary collaborations it could foster.
Collaborators include Alexander Stessin from Radiation Oncology; Wei Zhao specializing in breast cancer screening within Radiology; Haibin Ling from Computer Science at CEAS.