Figure 1.

Reconstruction of Mouse Poses and Quantification of Postures and Movements

(A) Body landmarks are separately tracked across each camera. See Figure S1 for additional details on the 4-camera system.

(B) A raw 3D reconstruction is obtained by triangulation of body landmark positions (left panel). The raw reconstruction is corrected by applying our algorithm based on the statistical shape model as described in STAR Methods, Figure S2, and Videos S1 and S2. The refined 3D reconstruction (right panel) is then used for all the further analyses.

(C) The model expressed by Equation 1 allows for quantifying a wide range of postures and movements of which red and green boxes report some examples. The “body shape” components enable to measure changes in body shape, such as body elongation and body bending. For additional details, see STAR Methods section “Interpretation of the eigenposes” and Figure S3. The “body position” components enable to quantify translations and rotations in a 3D space.

(D) The full set of behavioral measures, divided into 3 postural measures and 6 movement measures, is expressed as function of the terms in Equation 1. For additional details, see STAR Methods section “Validation of the postural and movement measures” and Figure S4.