Shape and texture prior for 2D/3D analysis/ synthesis of neural stem - - PowerPoint PPT Presentation

www.ipal.cnrs.fr

• A French-Singaporean joint laboratory

Shape and texture prior for 2D/3D analysis/ synthesis of neural stem cells for fate prediction 2D differentiation and 3D reconstruction

⎋

NEUROSPHERE FORMATION ASSAY

• Culture and purification “in vitro” of neural stem cell [1].
• Identify “proliferation” and “multi-potency”.
• No tracks of the origin cell & low survivability.
• Three dimension proliferation & two dimension differentiation.

2

Neuron Astrocyte Oligodendrocyte Stem Cell Progenitor Cell Neurosphere Growth Differentiation of Progenitor Cell Embriotic Mouse Brain

[1] S. Ahmed. The culture of neural stem cells. Journal of cellular biochemistery, 2009.

⎋

DATA

• Phase contrast time lapse sequence of 2D images.
• 40x magnification with 3x3 binning.
• 464 x 346 spatial resolution & 20 min time resolution.
• Online microscope tracking [2]

3 [2] C.-H. Huang, S. Sankaran, D. Racoceanu, S. Hariharan, and S. Ahmed. Online 3-d tracking of suspension cells imaged with phase-contrast microscopy. IEEE Transactions on Biomedical Engineering, 2012.

⎋

2D-3D REGISTRATION REVIEW

• Review: [3]
• Dimensionality problematic: projection, back-projection &

reconstruction.

4

back- projection 2D silhouette 3D model 2D silhouette 3D model projection 2D silhouettes 3D model 3D reconstruct model

[3] P . Markelj, D. Tomăzevĭc, B. Likar, and F. Pernŭs. A review of 3d/2d registration methods for image-guided interventions. Medical Image Analysis, 2012.

⎋

PROPOSED METHOD

• Analysis - cell detection
• Synthesis - model generation
• Registration / Optimisation - rank & optimise selected model

5 (1) Analysis (2) Synthesis (3) Registration - Optimisation

M i c r

• s

c

• p

e I m a g e

Registration Projection stop criterion

• ptimisation

3D Models Generation 2D Cell Detection

Convergence loop B e s t m

• d

e l s

Observation 2D data 3D data projected data

⎋

CELL DETECTION IN CLUSTER

6

ROI Level Set Hough Partial Circle Detection Local Maxima Detection Input Image Output Detection

⎋

2D IMAGE - 3D MODEL REGISTRATION

• Criteria

7

Shape Texture Topology Observation Model Registration

Feature based Intensity based

r

(xc,yc) (xc,yc)

• r

r 1

Texture modelisation

⎋

PROPOSED METHOD

8

M 1 M 2 M i E 1 E 2 E i Observation t Projection 1 Projection 2 Projection i

Score

⎋

RESULTS

9

Raw Temporary Detection results Registration error Model selected

⎋

CELL DETECTION IN CLUSTER

10

VIDEO

⎋

FUTURE WORK

• Model generation
• Integrate a dynamic generation.

11

Model Probability New Cell

⎋

NEXT STEPS

• From animal to human stem cells (skin)
• A suite of integrated microscopy systems for imaging anatomies of

complex 3D+T (4D) cell culture systems

• Drug testing platform for biologists using high-throughput imaging
• Next generation of neurodegenerative diseases treatments

12

⎋ International joint research unit - UMI CNRS 2955 www.ipal.cnrs.fr

Image & Pervasive Access Lab