Motivation

1
Motivation

• Microtubules (MT) are filamentous cytoskeleton
  structures composed of tibulin protein subunits.
• Quantitative analysis of MT dynamics in live cell
  is necessary
• Growth and shortening events of MT by considering
  only the MT tip position is not a valid
  approximation if MT follow non-linear paths
• Reliable tracing of MT body is required

2
Challenges

• Intersecting and overlapping MT regions appear
  brighter due to additive florescence
• Typical binarization methods yields gaps

3
Tip Detection

• MT image is filtered by second derivative of
  Gaussian filter with different orientations to
  reveal curvilinear structures.
• Binary mask obtained by thresholding the filter
  response is then further skeletonized to have
  B(x,y)
• The candidate tip positions are marked by finding
  the line ends in B(x,y)

4
Methodology

• Use estimated tip positions as starting points
• Apply fast marching to extract all possible paths
  starting with tip position and end points s.t.
• Calculate the support of the path p
• Select the path that maximizes the support

5
Results
Results of the Geodesic Paths
Results of the Proposed Algorithm