Fig. 6: Population decoding of audiovisual delays is enhanced in the posteromedial SC.
From: Functional specialisation of multisensory temporal integration in the mouse superior colliculus
A Anatomical distribution of individual recordings colour coded according to the mean AV delay classifier accuracy obtained with 25 neurons, 20 repetitions. B Decoding gradient as a function of ML and AP location, calculated from the results shown in (A). C Percentage of visual (V), auditory (A), delay-selective (D) and multisensory non delay-selective (M-D) neurons across four anatomical bins: AM anterior-medial, PM posterior-medial, AL anterior-lateral, PL posterior-lateral. Each bin covered 2.20 mm3, with on average 1039 ± 121 neurons per bin. D Percentage of neurons selective for different AV delays (0–100 ms) for the same anatomical bins as in (C). E AV delay decoding accuracy for 20 iterations of the linear classifier trained on the responses of 50 randomly sampled neurons within 650 μm bins along the ML (left), AP (centre) and DV (right) axes of the SC. The classifiers were either trained on the observed responses (black) or on temporally shifted linearly summed audio visual responses (red). The dotted line represents chance level. **P = 10−6; *P = 10−4 with respect to chance, two-sided Wilcoxon signed-rank test. P values indicate differences across anatomical locations for observed responses, two-sided Mann–Whitney U test. An 80% training and 20% testing split with 5-fold cross-validation was applied across all SVM implementations. The analysis of decoding accuracy corresponds to the test set.
