Talk outline • Long-range horizontal network and the association field • Electrophysiology, optical imaging : dynamics of propagating waves • A hypothesis • Non invasive approach in human : Perception and MEG • Functional role : reduced spike time variance • Spike Time Aligment Model : STAM)
The -static- Association field : a multidiciplinary approach Psychophysics Modelling Stemmler et al 1995, Somers et al. 1998, Field et al 1993; Hess & Dragoi et al 2000 Dakin 1997; Hess et al Series et al., 2001 1998; Kovacs & Julesz 1993,1994 Anatomy Physiology 80 80 Spikes/s Mc Guire et al. 1991 Gilbert & Wiesel, 1989 Gilbert, 1992 Sincich & Blasdel, 2001 Kapadia et al., 1995, 2000 Schmidt & al, 1997 Polat & al., 1998
Slow propagation wave of activity through intrinsic horizontal connections 0.1 to 0.5 m/s Space-time map Iso-latency map Optical recordings in Monkey V1 Intracellular recordings in Cat area 17 Grinvald & al., 1994; Glaser et al., 1999 Bringuier et al., Science, 1999
The stimulus-evoked population response in visual cortex of awake monkey is a propagating wave Lyle Muller, Alexandre Reynaud, Frédéric Chavane & Alain Destexhe ( 2014 ) Voltage Sensitive Dye (VSD) imaging
A Hypothesis Waves propagating through long range facilitation from “fast” neurons decreases the response latency of “slow” neurons (Fast/slow : high/low contrast ; magno/parvo cells) 200 Latency (ms) 100 0 0 50 100 Contrast Maunsell & al. (1999) As a consequence the variance of the response latencies at the population level is reduced –and the overall population response latency is shortened. Decreased variance and overall shortened latencies increases the firing probability of neurons at later stages and provides the population with a phase advance, and a temporal mechanism for figure/ground segregation
Hypothesis t1 t2 Firing threshold Cell 1 Resting state Propagating facilitation Cell 2 L1 L2 Using apparent motion to decompose and measure spatial facilitation dynamics Fast/slow -> Early/late
Experiments Using fast apparent motion to decompose and measure spatial facilitation dynamics Fast/slow -> Early/late Psychophysics: Perceptual correlates: apparent speed ? Magneto-encephalography: Correlates of propagating waves ? ┴ ┴ // // High contrast (50%) Low contrast (20%)
Perceptual correlate of propagating waves: apparent speed of fast apparent motion Gabor patches // path Gabor patches ┴ path 100 100 % Test faster Test vs. Reference speed Low / High Contrasts Low / Low 50 50 High / High High / Low 0 0 0.6 0.8 1 1.2 1.4 0.6 0.8 1 1.2 1.4 Speed ratio Psychophysics •8 participants •Reference speed: 60°/s •Test speeds: 36, 48, 60, 72 and 84°/s •2IFC Task: I1 or I2 faster ? •Conditions: 4 Contrast pairs x 2 Orientations x 5 Speed ratios •20 trials per condition•2 Temporal Alternative Forced Choice
The speeding-up effect: interpretation Result : Low contrast apparent motion sequences perceived faster than high contrast sequences, BUT only for parallel sequences t1 t2 200 Latency (ms) 100 Firing threshold Cell 1 Stimulus X1 T1 X2 T2 (T2=T1+dt physical ) Resting state 0 Propagating facilitation C ell V1 0 50 100 2 L1 L2 Contrast High contrast L high 1 short latency L high 2 = L high 1 +dt short latency Versus Low contrast L low 1 long latency L low 2 =L low 1 +dt long latency BUT DECREASED by spatial facilitation => physiological T1-T2 delay, dt apparent , shorter than physical dt ! MT : Speed = dX/dt apparent Perceptual decision : low contrast sequences appear faster Georges et al. 2001 Seriès et al 2001 Arnal & Lorenceau, unpublished data
Magneto-encephalography: Correlates of propagating waves ? Magneto-encephalography (MEG) : • 10 participants (2 excluded for bad signals) • Passive fixation • Conditions: 2 Contrasts x 2 Gabor orientations x 2 hemifields • 112 trials per condition • 151 axial magnetometers • Planar transformation • Individual sensor selection • Non-parametric tests on amplitude, peak latency and half-height latency (p-values presented for Wilcoxon tests) Fixation Fixation ISI 550-850 ms 0 533 ms 1300-1600 ms C High ┴ High // Low ┴ Low // Amplitudes Latency High contrast (50%) Low contrast (20%)
Magneto-encephalography: Individual sensor selection S1 (140 ms) S3 (180 ms) S4 (180 ms) S5 (190 ms) S6 (170 ms) S7 (180 ms) S8 (160 ms) S9 (160 ms) 200 fT 0 Stimulus: High ┴ S2 (140 ms) S10 (160 ms)
Magneto-encephalography: Results 17 ms Paradis et al. 2012
Magneto-encephalography: Source reconstruction Minimum Norm approach Source localization compatible with a V1 origin
Interim conclusions Perceptual and MEG results point to a significant influence of orientation and contrast on neural dynamics Phase advance for low contrast aligned Gabor patches (~17 ms) similar to electrophysiological results Functional role, if any ? Motion processing ? Eye-movement planning ? Contour processing ?
Motion processing ? • Speeding-up effect only seen for high speeds (~60°/s.) , rarely encountered in a natural environment (except during retinal slip induced by saccadic eye-movement). • Speeding-up effect only seen for low contrast apparent motion and aligned elements, unlikely to contribute to motion interpretation • Speeding-up effect: a side effect of neural dynamics developing within intrinsic long-range connections?
Eye movements – speeding-up saccadic- planning ? A tentative experiment (Arnal & Lorenceau, unpublished data) Collinear Orthogonal 1 1 0 0 -1 -1 -2 -2 -3 -3 -4 -4 -5 -5 -6 -6 -7 -7 -500 -400 -300 -200 -100 0 100 200 300 400 500 -500 -400 -300 -200 -100 0 100 200 300 400 500 Target
Eye movements – speeding-up saccadic- planning ? A tentative experiment (Arnal & Lorenceau, unpublished data) Mean Latencies ms. Duration ms. 179 27 178 177 26 176 175 174 Latency Saccade duration 25 173 285 * Saccade gain % 108 Max speed °/s 284 283 107 282 106 281 280 105 Saccade gain Maximum Speed 279 278 104 Orthogonal Colinear Orthogonal Colinear Gabor Orientation Gabor Orientation -> No significant evidence of speeded saccadic eye-movements -> Significant gain may reflect position uncertainty for aligned elements (inducing overshoot) Saccadic generation mostly in Superior Colliculus, lacking orientation selectivity
Spatial facilitation and Spike-Time Alignment Model STAM Fact: Contrast greatly varies along contours in natural images, implying an important latency variance at the population level, at odd with efficient integration of contours at later stages (e.g. from V1 to V2) Proposal: spatial facilitation between neurons responding to elongated contours with varying contrast favor (more) synchronized spiking
STAM: Chain of integrate-and-fire neurons connected (or not) by long-range facilitatory connections
Testing STAM with the border of Lena’s hat
Testing STAM with Lena’s hat A B C D Contrast distribution along Lena’s hat Paradis et al. 2012
Conclusions STAM implements a biologically plausible spatial facilitation dynamics between neurons responding to contours of varying contrast, resulting in spike time alignment (reduced latency variance at the population level). Reduced latency variance, and the resulting overall phase advance, of the neuronal population processing elongated contours in V1 (relative to neurons responding to “noisy” background) increases the probability, and reduces the latency, of firing at later processing stages, thus providing a temporal mechanism for figure/ground segregation. Earlier and more reliable responses of neurons at later stages (e.g. V2) that feedback onto V1 could in turn enhance the temporal contrast between contours and background
Recommend
More recommend
