Fibrillar Collagen Quantification With Curvelet Transform Based Computational Methods

Area of Science:

• Biomedical Engineering
• Computational Biology
• Medical Imaging

Background:

• Fibrillar collagen organization is crucial in diseases like breast and pancreatic cancer.
• Image-based biomarkers for collagen topology are needed for disease diagnosis and prognosis.
• Existing methods for collagen quantification lack advanced features for detailed analysis.

Purpose of the Study:

• To present an enhanced software platform for collagen quantification based on the curvelet transform (CT).
• To introduce new features for improved collagen fiber analysis, visualization, and validation.
• To demonstrate the software's utility in cancer pathology and preclinical models.

Main Methods:

• Utilized the curvelet transform (CT) algorithm for multiscale image representation, denoising, and edge enhancement.
• Developed C++ code optimization for rapid individual fiber tracking and a Java-based synthetic fiber generator for validation.
• Implemented automatic tumor boundary generation, parallel computing for batch processing, and region-of-interest analysis.

Main Results:

• Validated individual fiber tracking and orientation accuracy using synthesized data.
• Compared fiber orientation calculations with existing methods on pancreatic tissue, showing competitive performance.
• Demonstrated automatic tumor boundary creation and collagen fiber alignment quantification in human breast cancer and in vivo mouse xenograft models.

Conclusions:

• The CT-based software platform offers a robust and efficient solution for quantifying collagen organization.
• New features enhance accuracy, speed, and applicability in diverse cancer research settings.
• The tool shows significant potential for identifying novel image-based biomarkers in cancer diagnostics.