Quantitative Analysis of TDLUs using Adaptive Morphological Shape Techniques.

Area of Science:

• Biomedical Engineering
• Computational Pathology
• Breast Cancer Research

Background:

• Terminal duct lobular units (TDLUs) are the primary origin of breast cancer.
• TDLUs undergo age-related involution, and lack of this process is linked to increased cancer risk.
• Current TDLU assessment relies on subjective visual evaluation.

Purpose of the Study:

• To develop an automated system for quantitative measurement and classification of TDLUs.
• To provide objective descriptors for TDLUs, aiding in breast cancer risk assessment.
• To explore novel structural measures of acini within TDLUs using machine learning.

Main Methods:

• Development of a computational technique for automatic quantitative analysis of TDLU morphology.
• Validation against pathologist assessment using breast tissue from the Susan G. Komen Tissue Bank (n=51).
• Application to a clinical dataset from the National Cancer Institute's BREAST Stamp Project (n=52) for broader applicability.
• Utilizing machine learning and clustering to identify novel structural properties.

Main Results:

• The automated system achieved 70% agreement with pathologist assessment on research tissues.
• An 82% correlation with visual assessment was obtained on clinical biopsy samples.
• Machine learning revealed that acini number increases exponentially with TDLU diameter, while elongation and roundness remain constant.

Conclusions:

• Automated quantitative analysis of TDLUs offers an objective alternative to subjective visual assessment.
• The developed method demonstrates accuracy and applicability across different datasets.
• This approach can uncover novel insights into TDLU structural properties relevant to breast cancer research.