Longitudinal case-control study of mammographic breast tissue subtypes
When |
Jul 03, 2023
from 11:00 to 12:00 |
---|---|
Where | Salle des Thèses |
Attendees |
André Khalil |
Add event to calendar |
vCal iCal |
Breast cancer is the most diagnosed cancer globally. Numerous deep learning and other artificial intelligence (AI) techniques are being proposed to help with routine screening mammography exams. However, the lack of interpretability of these AI techniques causes a reluctance of acceptance by clinicians. Additionally, they require extensive training and their generalizability has yet to be robustly demonstrated. For example, recall rates may increase substantially (up to 3-fold) following mammography equipment software upgrades. Coincidentally, too many women are recalled for diagnostic imaging, leading to unnecessary emotional, temporal, and financial burdens.
Mammographic percent density (MPD), which is quantified on a woman’s annual screening mammograms, is an independent risk factor for breast cancer. However, integrating MPD and polygenetic scores into traditional risk models only marginally increases their accuracy. More efficient strategies are needed to identify the at-risk subset of women with mammographically dense tissue. There has been limited research on potential subtypes of mammographic dense breast tissue to identify areas of risky dense tissue that is structurally unorganized and links to cancer dynamics, versus areas of healthy dense tissue which remains organized. We hypothesize these subtypes of mammographic dense tissue may provide insights into breast cancer risk. A retrospective age-matched study was conducted to investigate breast cancer using longitudinal screening mammograms from 25 benign cases, 21 cancer cases, and 24 controls. By measuring the area of each tissue type using a patented approach, we found that the amount of risky and healthy dense tissue, and the rate of change in risky dense tissue was associated with developing breast cancer. Moving forward, our goal is to establish risky dense tissue as a better predictor of breast cancer by showing risky dense tissue is more accurate in predicting breast cancer than MPD on a retrospective study of 9,000 patients. Further validation will help medical professionals proactively stratify risk and recommend preventive care rather than reactive treatments to improve patient outcomes.