Neuron Tracking In Calcium Image Stacks Using Accordion Articulations

The output of several distinct classes of proprioceptive sensory neurons in the Drosophila larva is required for coordinated locomotion. Yet how these neurons respond to idio-thetic movements remains unknown. The lack of robust tools for simultaneous quantification of both morphology and neuronal activity in calcium images hinders the progress. The main bottleneck for tracking deformations is a combination of (a) imaging-specific attributes, namely low axial (Z) resolution and noise, and (b) movement-specific characteristics, namely strong deformations yielding a strain that travels as a wave from one side of the ventral nerve cord segment to the other. Fortunately, the wave-like form of forward and backwards locomotion allows for model constraints that compensate for the imaging ambiguities. In this work, we developed a method for tracking sensory neurons in the Drosophila larva, using appearance, local shape constraints, and global morphology characteristics embedded in an accordion-like, piece-wise articulated surface that approximates local maximum average intensity and incorporates the locomotion characteristics. Our results show that this approach can accurately track local branch deformations, but most importantly, it can estimate dendrite axial deformations, i.e., bending curvatures along the Z-axis, under the progressive strain wave during locomotion.