Collagen organization plays an integral role in many diseases including cancer. Here we introduce a low-cost, open-access collagen imaging and image analysis platform for quantifying fibrillar collagen organization. LC-PolScope was used as the imaging modality that incorporates a precision universal compensator made of two computer controlled liquid crystal variable retarders. This imaging system can be easily implemented on standard microscopes as a cost-effective alternative to second harmonic generation (SHG) imaging and staining on a wide range of available pathology slide formats, including the most commonly used H&E stained slides. In the collagen image analysis, a two-step registration process was first used for overlaying bright-field images on polarized images of collagen: 1) Extract collagenous stroma from H&E bright-field images by image segmentation in HSV color space and performing color separation using K-means clustering algorithm to find the best collagen estimate; 2) Use an iterative intensity-based image registration algorithm to find the affine transform that registers the collagen extracted image to the SHG image at different resolutions. Then, the registered bright field H&E image was used as a guide to evaluate collagen organization near any biological structure such as blood vessels, tumors etc. These algorithms have been implemented in our open source collagen analysis software tool “CurveAlign” package that has been widely used for collagen feature extraction, including detection of tumor associated collagen signatures. As a proof of concept, we are now using this platform to investigate collagen organization in metastatic pancreatic cancer vs non-metastatic pancreatic cancer.
|