CurveAlign for Fibrillar Collagen Quantification (Current version-V5.0 available on GitHub)

CurveAlign  is a quantitative tool for interpreting the regional interaction between collagen and tumors by assessment of up to ~thirty stromal fiber features, including angle, alignment, and density. CurveAlign can incorporate fiber extraction from CT-FIRE, another open source tool best suited for individual fiber analysis. We have been actively developing new modules to incorporating individual cells information, improving fiber tracking in complex network organizations, providing interfaces to other open-source tools. We are particularly interested in the implementation of state-of-the-art machine learning based methods to quantifying interactions between collagen, tumor cells, and immune cells.

 When do I use CurveAlign and when do I use CT-FIRE?

These two programs were developed with complementary but slightly different main goals. CurveAlign (Schneider et al., Bredfeldt et al., 2014a;  Liu et al., 2017; Liu et al., 2020) was developed first and had the main goal of quantifying all fiber angles within a region of interest relative to a user defined boundary be it a straight line or a tumor boundary. As our research grew in investigating the role of collagen in cancer progression and invasion we wanted to investigate how individual fiber parameters could influence cancer and other diseases. Out of this need came the development of CT-FIRE(Bredfeldt et al., 2014b) to analyze individual fiber metrics such as length, width, angle, and curvature.  Besides the relative angle quantification with respect to the boundary, the current version of CurveAlign can be used to extract other collagen fiber features, such as localized fiber density, fiber alignment, and the spatial relationship between fiber and the associated boundary. In addition, the extracted individual fibers extracted by CT-FIRE can be imported into the CurveAlign for additional feature extraction mentioned above. We have future plans to integrate these programs further. For now, CurveAlign should be used for bulk assessment of collagen features including alignment/density and CT-FIRE for individual fiber quantification.

 What can CurveAlign do?

The purpose of CurveAlign is to compute global and localized collagen features that describe collagen remodeling due to its interactions with epithelial cells or other causes. It was developed in order to search for stromal changes that are correlated with disease in images of collagen and epithelial cells. The primary major features may include the angle related to the defined boundary, localized the density and alignment, etc.  Either curvelets in the Curvelet transform or the segmented fibers in CT-FIRE can be used to represent the actual fiber organization depending on the application conditions. The output may be displayed on the screen and written into .csv, .xlsx, .mat data file or .tif image files for further analysis, such as statistical test or image classification.

CurveAlign 5.0 demo
CurveAlign main GUI, output table, output histogram, overlay figure, and heatmap of orientations.

Key features include:

  • Emphasis is on collagen fiber feature extraction with respect to tumor boundary and nearby fibers
  • Analysis of many types of microscopy data including Second Harmonic, Multiphoton and histology stains
  • Measure global orientation related metrics
  • Use nearest-neighbors algorithm to measure localized orientation, alignment, and density of collagen fibers
  • Use region of interest analysis module for localized fiber feature extraction
  • Use curvelets in the curvelet transform to represent collagen fibers
  • Import the fibers extracted by CT-FIRE to represent collagen fibers
  • Use linear SVM to rank the importance of fiber features in discriminating images from different groups

 Main release notes:

  CurveAlign v5.0 Beta major updates ( July, 2020):

  • Add another option for the registration (in HSV color space) of  H&E bright field image and SHG image pair
  • Add density analysis module in CurveAlign ROI manager
  • Add faster individual fiber estimation in CT-FIRE module
  • Add java based synthetic fiber generator
  • Add headless option to run on cloud systems (such as CHTC at UW-Madison)
  • Add alignment calculation for fibers within the ROI determined by the distance to the boundary
  • Add an optional minimum distance check to remove fibers on or very close to a boundary

    CurveAlign v4.0 Beta major updates (August, 2017):

  • Add an automatic boundary creation module to automatically  register H&E image with SHG image, and then segment boundaries based on the registered H&E bright field image
  • Add the ROI manager for defining the regions of interest in an image
  • Add ROI analysis for post-processing CurveAlign analysis results of an individual image
  • Add ROI analysis for running CurveAlign on the defined ROIs of individual image for orientation and alignment measurement with Curvelets
  • Add ROI analysis and ROI post-processing for the defined ROIs in multiple images
  • Add output table to selectively display the alignment information as well as  the associated overlay, heatmap and histogram figures
  • Add options in Curvelets-based analysis to allow the user to specify the scale to be used and the radius to group the curvelets
  • Add an option to select/un-select the fibers within a tiff boundary
  • Add options to control the heat map of the relative angle or localized alignment
  • Add a function to manually draw tiff boundaries
  • Fix a bug in the measurement and output of the relative angle
  • Add the function to combine the results in a selected folder of CurveAlign output

 CurveAlign version 3.0 major updates (December 2014):

 Compared to the version 2.3, its primary change is adding fiber feature extraction from CT-FIRE output

  • Output up to thirty four fiber features, including angle, alignment, density, and etc. , these features are saved in both .csv file and .mat file
  • Multiple tiff boundaries can be loaded to investigate the interaction between fiber and cell
  • Read in CT-FIRE fiber extraction results
  • A feature ranking using linear support vector machine approach is included
  • Keep the most features of the older version (2.3), such as the statistics for the relative angles, heat map to show the fiber (angle) alignment, and the fiber-boundary association, etc.

Frequently Asked Questions about CurveAlign can be seen here.


Yuming Liu (primary contact and lead developer, Aug 2014-)

Tim Liang (Undergraduate student developer, Jan 2023-)

Nathan Labiosa (Undergraduate student developer, Feb 2023-)

Michael Nelson(PhD student, June 2021-)

Helen Wilson (PhD student, Jan 2022-)

Bin Li (Collaborator and former PhD student, Oct 2019-)

Hyojoon Park (“Joon”, PhD student, March 2022-)

Previous Main Contributing Developers:

Heqiao (“Wonder”, Undergraduate student developer, Aug 2020-May 2022)

Wenxin (“Sabrina”, Undergraduate student developer, May 2021-May 2022)

Matthew Dutson (PhD student developer, Sep 2018-Sep 2019)

Akhil Patel (Undergraduate student developer, May 2019-Sept 2019)

Adib Keikhosravi (PhD student developer, Aug 2014-March 2019)

Michael Pinkert (PhD student developer, May 2017-March 2019)

Haixiang Liu (PhD student developer, Aug 2018- Feb 2019)

Robert Claus (Graduate student developer, Aug 2018-Oct 2018)

Andrew Leicht (former student developer, Aug 2017-Dec 2017)

Guneet Singh Mehta (MS student developer, Aug 2014-May 2017)

Jeremy Bredfeldt (PhD student developer, Jun 2012-  Jul 2014)

Carolyn Pehlke (PhD student developer, Jan 2009- May 2012)

Downloads (latest version, CurveAlign V5.0 Beta) 

Standalone package for Windows 64 bit

Standalone package for MAC

MATLAB m-files

Manual and testing images

Downloads for all available versions


  • Matlab standalone APP: See the manual included in the package above
  • MATLAB source code version:Download and unzip the Matlab m-files above. Then go to and register to sign a licensing agreement and download the CurveLab 2.1.2 MATLAB package. Place the folder “CurveLab-2.1.2” into the CurveAlign folder. With MATLAB’s Current Folder set to the CurveAlign folder, enter “CurveAlign” at the command prompt to launch the GUI.

Source Code: latest code at GitHub

Website:  GitHub wiki page

Citing CurveAlign

Original implementation: Bredfeldt, J.S., Liu, Y., Conklin, M.W., Keely, P.J., Mackie, T.R., and Eliceiri, K.W. (2014b). Automated quantification of aligned collagen for human breast carcinoma prognosis. J. Pathol. Inform. 5. PMID: 25250186

Protocol of using CurveAlign+CT-FIRE : Liu, Y., Keikhosravi, A., Mehta, G.S., Drifka, C.R., and Eliceiri, K.W. (2017). Methods for quantifying fibriliar collagen alignment. In Fibrosis: Methods and Protocols, L. Rittié, ed. (New York: Springer). PMID: 28836218


Bredfeldt, J.S., Liu, Y., Pehlke, C.A., Conklin, M.W., Szulczewski, J.M., Inman, D.R., Keely, P.J., Nowak, R.D., Mackie, T.R., and Eliceiri, K.W. (2014a). Computational segmentation of collagen fibers from second-harmonic generation images of breast cancer. J. Biomed. Opt. 19, 016007–016007. PMID: 24407500

Bredfeldt, J.S., Liu, Y., Conklin, M.W., Keely, P.J., Mackie, T.R., and Eliceiri, K.W. (2014b). Automated quantification of aligned collagen for human breast carcinoma prognosis. J. Pathol. Inform. 5. PMID: 25250186

Liu, Y., Keikhosravi, A., Mehta, G.S., Drifka, C.R., and Eliceiri, K.W. (2017). Methods for quantifying fibriliar collagen alignment. In Fibrosis: Methods and Protocols, L. Rittié, ed. (New York: Springer). PMID: 28836218

Liu, Y., Keikhosravi, A., Pehlke, C. A., Bredfeldt, J. S., Dutson, M., Liu, H., Mehta, G. S., Claus, R., Patel, A. J., Conklin, M. W., Inman, D. R., Provenzano, P. P., Sifakis, E., Patel, J. M. & Eliceiri, K. W. (2020). Fibrillar collagen quantification with curvelet transform based computational methods. Front. Bioeng. Biotechnol. 8, 198. PMID: 32373594

Schneider, C.A., Pehlke, C.A., Tilbury, K., Sullivan, R., Eliceiri, K.W., and Keely, P.J. (2013). Quantitative Approaches for Studying the Role of Collagen in Breast Cancer Invasion and Progression. In Second Harmonic Generation Imaging, F.S. Pavone, and P.J. Campagnola, eds. (New York: CRC Press), p. 373.