Shannon Quinn - Python for Public Health: Building Statistical Models of Ciliary Motion

Cilia, microscopic hairs lining nearly every cell surface in your body, play a major role in developmental and sinopulmonary health. As such, deriving quantitative properties of their motion is compelling for both clinical and research purposes. Here we demonstrate a computational pipeline built entirely in Python for analyzing ciliary motion using a variety of machine learning techniques.

https://us.pycon.org/2016/schedule/presentation/1580/

Transcript

1. Python For Public Health:
   Building Statistical Models
   Of Ciliary Motion
   Shannon Quinn
   University of Georgia
   PyCon 2016
   

   View Slide

2. Part I: Cilia and their pathologies
   

   View Slide

3. What are cilia?
   Scale bars: 10µm
   

   View Slide

4. Why do we care about cilia?
   u Clinical
   u Ciliopathies
   u Association with
   congenital heart disease
   u Developmental
   u Nodal flow
   u Left-right asymmetry
   

   View Slide

5. How do we diagnose ciliopathies?
   Cheap, fast, inaccurate Slow, expensive, accurate (?)
   Measure nasal
   nitric oxide
   (NO) levels
   Electron
   microscopy to
   search for
   structural
   defects
   Ciliary beat
   frequency
   (CBF)
   computation
   Manual ciliary
   beat pattern
   analysis
   “Gold standard”
   

   View Slide

6. What is our goal?
   u Input: high-speed video of ciliary biopsy
   u Output: quantitative properties of observed motion
   Curly!
   

   View Slide

7. Part II: Quantitative features of
   ciliary motion
   

   View Slide

8. Strategy for quantifying motion
   

   View Slide

9. From videos to features
   

   View Slide

10. Features describing motion
    Scaling (zoom) Deformation
    (biaxial shear)
    Rotation (curl)
    Not useful in
    2D
    “Novel use of differential image velocity invariants to categorize ciliary
    motion defects.” Quinn SP, Francis R, Lo C, Chennubhotla CS. Proceedings
    of the Biomedical Science and Engineering Conference (BSEC) 2011.
    

    View Slide

11. What do these features look like?
    Rotation (rad/s)
    

    View Slide

12. How do we represent these
    features?
    ~
    yt =
    C~
    xt
    ~
    xt =
    A1~
    xt 1 +
    A2~
    xt 2 +
    ...
    +
    Ad~
    xt d
    Feature
    vectors!
    

    View Slide

13. What can we do with these
    features?
    93%
    classification
    accuracy
    Automated identification of abnormal respiratory
    ciliary motion in nasal biopsies. Quinn SP, Zahid
    M, Durkin J, Francis R, Lo C, Chennubhotla CS.
    Science Translational Medicine 2015.
    

    View Slide

14. Part III: The Python ecosystem
    

    View Slide

15. Collecting data from collaborators
    1. Upload to django website
    2. Annotate manually
    3. Save annotations in
    database using SQLAlchemy
    

    View Slide

16. Processing raw video data
    1. Videos in AVI format
    2. Convert to NumPy arrays
    with OpenCV
    3. Compute optical flow with
    OpenCV
    4. Compute rotation with SciPy
    signal processing filters
    

    View Slide

17. Derivation of time series models
    u …custom code!
    u scipy.linalg
    

    View Slide

18. Classification pipeline
    u Cross-validation for model
    fitting and testing using
    scikit-learn
    u joblib / PySpark for
    parameter scanning
    u matplotlib + seaborn /
    bokeh for visualization
    

    View Slide

19. (BONUS ROUND)
    Part IV: Conclusions and future
    directions
    

    View Slide

20. Conclusions
    u 93% classification: methods are
    sound
    u Dynamic texture representation is
    accurate
    u Blackbox tool for clinicians
    u Web front-end + Python
    middleware + Spark back-end
    u Upload video -> Get analysis
    u Assist experts with diagnostics
    u Expert input
    u Phenotype annotations, regions of
    interest
    

    View Slide

21. Future directions
    u Normal / Abnormal is an
    oversimplification…
    u Unsupervised motion
    clustering pipeline
    u Associate specific motion
    phenotypes with
    clinical conditions
    

    View Slide

22. Thank you!
    u spq@uga.edu
    u @SpectralFilter
    u https://magsol.github.io/
    

    View Slide