Multiple-target tracking of 3D fluorescent objects based on simulated annealing

Victor Racine (1) , Ariane Hertzog (1) , Jacqueline Jouanneau (1) , Jean Salamero (1) , Charles Kervrann (2, 3) , Jean-Baptiste Sibarita (1)

1 Institut Curie [Paris]
2 MIA-JOUY - Unité de biométrie et intelligence artificielle de Jouy
3 VISTA - Vision spatio-temporelle et active

This paper presents a framework for tracking fluorescent objects imaged with 3D+time video-microscopy. The proposed technique solves the NP-hard problem of the multi-frame object correspondence. For this purpose, we use the simulated annealing algorithm for its flexibility for feature correspondence and its ability to give results in a very short time. Our approach takes into account events like "split", 'merge", "birth" and "death". These "evolutionary events" are motivated by biological and physical considerations. We demonstrate the performance of the proposed algorithm on images of carcinoma NBT-II cells expressing fluorescent LAR-PTP, a tyrosine phosphatase possibly involved in tumor progression.

Mots clés: motion correspondence

Mots clés: data association; points; correspondance de mouvement; association de données

Domaines: Mathématiques [math]; Informatique [cs]

Victor Racine, Ariane Hertzog, Jacqueline Jouanneau, Jean Salamero, Charles Kervrann, et al.. Multiple-target tracking of 3D fluorescent objects based on simulated annealing. 3. IEEE International Symposium on Biomedical Imaging, Apr 2006, Arlington, United States. ⟨hal-02757151⟩