Automated MRI-based breast cancer segmentation

This technology is an automated, AI-based platform for identifying and segmenting ductal carcinoma in situ (DCIS) lesions on breast MRI. .

Unmet Need: High-throughput, reliable breast cancer imaging tools

Breast MRI is widely used for detecting and characterizing breast cancer, but accurate delineation of lesions– particularly ductal carcinoma in situ (DCIS)– remains challenging due to heterogeneous enhancement patterns and obscuration by dense tissue. Current workflows rely heavily on manual or semi-manual segmentation by expert radiologists, which is time-consuming, subjective, and difficult to scale across large imaging datasets. These limitations hinder consistent lesion characterization, longitudinal monitoring, and large-scale analysis. Addressing these issues is essential to improve reproducibility, reduce clinical burden, and develop future diagnostic and treatment tools for breast cancer.

The Technology: Scalable AI-driven breast MRI segmentation

This technology harnesses the power of a large-scale, pre-trained medical image segmentation model to deliver robust, reliable breast cancer analysis from MRI. By adapting rich, generalizable image features learned across diverse imaging modalities to breast MRI, the system accurately identifies and delineates cancerous lesions despite variability in lesion presentation and breast tissue composition. Designed for seamless integration into clinical and research workflows, it operates with minimal manual input while ensuring consistent, reproducible results across datasets. Prototype evaluations on breast MRI demonstrate strong performance in detecting and segmenting invasive lesions, highlighting its potential to improve efficiency and confidence in breast cancer imaging analysis.

Applications:

  • Diagnostic tool to classify cancer disease state
  • Pre-operative planning to improve surgical margin estimation
  • Longitudinal disease tracking
  • Treatment response monitoring
  • Biomarker extraction
  • Decision support making tools
  • Clinical training tools

Advantages:

  • High-throughput platform for cancer imaging analysis
  • Unbiased platform, less prone to human error
  • Scalable and compatible with existing technology
  • Reduces reliance on manual segmentation, saving time and money
  • Handles heterogeneous lesion appearance
  • Improves consistency and reproducibility
  • Improves accessibility in resource-limited settings

Lead Inventor:

Despina Kontos, Ph.D.

Related Publications:

Tech Ventures Reference: