Description »

This project will require the student to develop an innovative user interface that allows biologists to interact with and correct segmentations extracted from neuronal data, such as the one shown in the above video. The data may be 2D, 3D or higher dimensional (3D time-lapse, or 3D color for instance). Interacting with high dimensional data presents unique challenges which will require innovative solutions. Furthermore, correcting the segmentations will require new methods to be developed, requiring the student to understand the Geodesic Tube method [1] and the k-MST algorithm [2], which are used to perform the segmentations. Some examples are shown below. Finally, we would like to investigate online learning methods, in which the user improves the segmentation by iteratively correcting and retraining the algorithm.

Original images and tubular tree segmentation. Visualized using Fiji.

References:

• [1] F. Benmansour and L. Cohen. Tubular Structure Segmentation based on Minimal Path Method and Anisotropic Enhancement. IJCV 2010.
• [2] E. Turetken, G. Gonzalez, C. Blum, and P. Fua. Automated Reconstruction of Dendritic and Axonal Trees by Global Optimization with Geometric Priors. Neuroinformatics, In Press, 2011.