MRI-based radiomic analysis for DCIS risk stratification and treatment planning

This technology is a computational imaging method that utilizes quantitative radiomic features extracted from standard-of-care breast MRI scans, combined with clinical information, to predict which patients with ductal carcinoma in situ (DCIS) are at risk for disease progression to invasive cancer.

Unmet Need: MRI-based computational analysis for personalized DCIS risk assessment

Ductal carcinoma in situ (DCIS) accounts for approximately 20% of breast cancer diagnoses, yet current clinical approaches cannot reliably distinguish low-risk from high-risk disease. This leads to overtreatment in up to half of diagnosed women, who undergo unnecessary surgery and radiation therapy with associated morbidity and side effects. Current qualitative MRI interpretation is sensitive for detecting DCIS, but it cannot identify which patients are at risk for upstaging to invasive disease at surgery. There is a critical need for improved risk stratification tools that can identify low-risk patients suitable for active surveillance and help match treatment intensity to tumor biology.

The Technology: Quantitative MRI radiomic phenotyping for predicting DCIS disease upstaging

This technology extracts high-throughput quantitative imaging features (radiomics) from standard dynamic contrast-enhanced (DCE) breast MRI scans to characterize ductal carcinoma in situ (DCIS) lesions. Using the publicly available Cancer Phenomics Toolkit (CaPTk), radiomic features are computed from lesions on the initial post-contrast MRI. Two complementary approaches are then applied: (1) unsupervised clustering identifies distinct phenotypes that correspond to low- and high-risk groups, and (2) principal component analysis reduces the feature space to key components capturing 85% of variance. These radiomic metrics are combined with standard clinical variables and qualitative MRI features in logistic regression models. The combined model achieved an AUC of 0.77 for predicting disease upstaging and identified 25% more true-negative cases compared to using clinical information alone at 88% sensitivity.

This technology has been validated in a multicenter clinical trial with 295 DCIS patients across diverse scanner types and imaging protocols.

Applications:

  • Risk stratification tool for selecting ductal carcinoma in situ (DCIS) patients for active surveillance programs
  • Clinical decision support system for guiding treatment de-escalation in low-risk DCIS
  • Preoperative assessment tool to predict the likelihood of upstaging to invasive disease
  • Personalized medicine platform for matching DCIS treatment intensity to tumor biology
  • Multicenter clinical trial endpoint for evaluating novel DCIS management strategies
  • Quality assurance tool for standardizing DCIS risk assessment across institutions

Advantages:

  • Utilizes standard-of-care DCE-MRI without requiring specialized imaging sequences
  • Provides quantitative, objective metrics that augment subjective radiologist interpretation
  • Demonstrates robustness across diverse scanner manufacturers
  • Achieves 96% negative predictive value for ruling out disease upstaging at surgery
  • Employs publicly available open-source software (CaPTk) for clinical translation
  • Enables non-invasive risk stratification without additional biopsies or procedures
  • Outperforms models using only clinical or qualitative MRI features (AUC 0.77 vs 0.72)

Lead Inventor:

Despina Kontos, Ph.D.

Related Publications:

Tech Ventures Reference: