Sandeep Bhat Systems Bioimaging Lab, UCSB Presentation for ENGR 103, Nov 20, 2008

Blind Source-Separation for Heart Wall and Erythrocyte Motion Estimation in the Beating Embryonic Heart

Sandeep BhatSystems Bioimaging Lab, UCSBPresentation for ENGR 103, Nov 20, 2008

sandeepkbhat@ece.ucsb.edu

2

Agenda

• Motivation• Problem Setup• Algorithm stages• Results• Conclusion

3

Motivation

• Study of model organisms

– Embryonic development, toxicology, specific gene function.

– Zebrafish embryo is transparent

• Live imaging of embryos.

– Fluorescence microscopy (FM)

– Bright field microscopy (BFM)

• The algorithm aids BFM imaging of developing zebrafish heart.

4

Problem Setup

• Input: BFM image sequence

– Spanning several cardiac cycles.

• Desired output: Two image sequences

– Spanning one period (cardiac cycle)

– One showing only the heart morphology

– The other with only the transient structures (Red Blood Cells)

5

Algorithm Stages

• Three stages:

– Pre-processing

– Separation algorithm

– Post-processing

6

Separation Algorithm

• Temporal alignment

• Heart-wall:

– “Periodic”

– Mean/Median filtering

• Red Blood Cells:

– Not periodic

– Subtraction/Variance

7

• Flow analysis using FlowJ

Post-processing

8

Results• frequency components. • Testing with synthetically generated data.• Testing with synthetically generated data.

• (a) Mean+Subtraction (b) Mean+Variance

• (c) Median+Subtraction (d) Median+Variance

9

Results (contd.)• BFM images of zebrafish heart

10

Conclusion