Refined movement analysis in the staircase test reveals differential motor deficits in mouse models of stroke

Abstract

Accurate assessment of post-stroke deficits is crucial in translational research. Recent advances in machine learning offer precise quantification of rodent motor behavior post-stroke, yet detecting lesion-specific upper extremity deficits remains unclear. Employing proximal middle cerebral artery occlusion (MCAO) and cortical photothrombosis (PT) in mice, we assessed post-stroke impairments via the Staircase test. Lesion locations were identified using 7 T-MRI. Machine learning was applied to reconstruct forepaw kinematic trajectories and feature analysis was achieved with MouseReach, a new data-processing toolbox. Lesion reconstructions pinpointed ischemic centers in the striatum (MCAO) and sensorimotor cortex (PT). Pellet retrieval alterations were observed, but were unrelated to overall stroke volume. Instead, forepaw slips and relative reaching success correlated with increasing cortical lesion size in both models. Striatal lesion size after MCAO was associated with prolonged reach durations that occurred with delayed symptom onset. Further analysis on the impact of selective serotonin reuptake inhibitors in the PT model revealed no clear treatment effects but replicated strong effect sizes of slips for post-stroke deficit detection. In summary, refined movement analysis unveiled specific deficits in two widely-used mouse stroke models, emphasizing the value of deep behavioral profiling in preclinical stroke research to enhance model validity for clinical translation.

Keywords: Machine learning; motor deficits; rodent models; stroke; translational research.

Figure 1.

Setup for refined movement analysis and validation of algorithm performance. (a) Four Staircase boxes are simultaneously recorded using two highspeed cameras that are triggered by infrared beams, when mice enter the reaching chambers. (b) Images from individual video-frames are segmented and streamlined for machine learning based motion tracking of forepaw and pellet trajectories in individual Staircase boxes. (c) The direction of forepaw movement during reach cycles defines movement towards (reach-phase) and away from a target sugar pellet (retraction-phase). (d) Successful and unsuccessful reaches are detected based on logical threshold operations for forepaw and pellet displacements. Slips were characterized by sudden vertical movements, independent of pellet information. Probability denotes the probability of pellet existence in a staircase well. Δx and Δy: horizontal and vertical forepaw displacement; y: vertical pellet position. (e) The validation of automatically detected reaches and slips shows high degrees of accuracy in comparison to visual annotation by blinded raters. Data in e present results for two mice from the MCAO and PT groups, before and after stroke on days 7, 14, and 21.

Figure 2.

Experimental timeline and quantification of lesion volumes and locations. (a) Animals were trained in the Staircase daily (green bar), except for the day of stroke surgery. Video recordings were performed on day 4 before, and days 7, 14, and 21 after stroke. Days before and after stroke are indicated with symbol ‘-’ and letter ‘p’. MRI was performed on day 1 after stroke. (b) Representative MRI cross sections after MCAO and PT (grey scale images), and colored stroke incidence maps at the group level for MCAO (n = 10) and PT (n = 9) mice. Color bar indicates percentage of mice that showed lesions in individual MRI voxels. (c) Total lesion volume and percentage of lesion affecting cortical, striatal, and residual brain areas. Percentages in c are referenced to the stroke affected hemisphere. Bar graphs are reported as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 3.

Lesion-specific deficits in the contra-lesional forepaw following MCAO or PT in mice. (a) Linear discriminant analysis (LDA) performed on 30 outcome parameters separates stroke groups and recording days. (b) LD1 scores show significant changes in post-stroke behavior for MCAO and PT groups. (c) The top ten factor loadings on LD1 reveal contribution of speed-related and global parameters to post-stroke deficits. (d) Comparison of post-stroke deficits using traditional quantification (pellet count) vs. refined outcome parameters. Bar graphs are reported as mean ± SD and box plots with Tukey method, *p < 0.05, **p < 0.01.

Figure 4.

Correlation between outcome parameters and either total lesion volume, cortical or striatal lesion percentage. Correlations are reported for pooled dataset of MCAO and PT animals on day 7 after stroke for traditional pellet count (a), success coefficient (b) and slip depth (c), as well as on day 21 for normalized duration of the reaching cycle (d). Lines indicate linear fit and 95% confidence intervals. Red and blue shaded areas in d indicate decreased or increased reaching duration in comparison to pre-stroke behavior. Values report Pearson coefficients (r) and corresponding p-values. Significant correlations are marked in bold font for p < 0.05.

Figure 5.

Compensatory kinematic changes in the ipsi-lesional forepaw. (a) Linear discriminant analysis (LDA) identifies the presence of ipsi-lesional motor adaptations, following MCAO and PT. (b) LD1 scores show significant changes on different recording days for both stroke groups. (c) Speed and distance-related parameters primarily account for the observed movement changes, as shown by the top 10 factor loadings on LD1. (d) Individual parameters provide information on new movement strategies after PT, composed of faster and shorter reaching movements. Bar graphs are reported as mean ± SD and box plots with Tukey method, *p < 0.05, **p < 0.01.

Figure 6.

Validating refined outcome parameters during SSRI treatment in the PT model. (a) Experimental design. We compared the effect of daily SSRI treatment with citalopram (10 mg/kg i.p.) to saline injections on the evolution of post-stroke deficits (n = 5 mice per group). Injections started two days after ischemia. Video recordings in the Staircase and MRI measurements were performed on the timepoints indicated in the illustration. (b) Comparison of main outcome parameters, including the number of reach events, traditional pellet count, success coefficient and slip events in their abilities to capture post-stroke deficits and treatment effects. Box plots are reported with Tukey method. *p < 0.05 for post-hoc group comparison with pre-stroke values for each group.