Yukako T. Hasegawa

Single trial-based prediction of a go/no-go decision in monkey superior colliculus

While some decision-making processes often result in the generation of an observable action, for example eye or limb movements, others may prevent actions and occur without an overt behavioral response. To understand how these decisions are made, one must look directly at their neuronal substrates. We trained two monkeys on a go/no-go task which requires a saccade to a peripheral cue stimulus (go) or maintenance of fixation (no-go). We performed binary regressions on the activity of single neurons in the superior colliculus (SC), with the go/no-go decision as a predictor variable, and constructed a virtual decision function (VDF) designed to provide a good estimation of decision content and its timing in a single trial decision process. Post hoc analyses by VDF correctly predicted the monkeys choice in more than 80% of trials. These results suggest that monitoring of SC activity has sufficient capacity to predict go/no-go decisions on a trial-by-trial basis.

Inhibition in superior colliculus neurons in a brightness discrimination task

Simultaneous recordings were collected from between two and four buildup neurons from the left and right superior colliculi in rhesus monkeys in a simple two-choice brightness discrimination task. The monkeys were required to move their eyes to one of two response targets to indicate their decision. Neurons were identified whose receptive fields were centered on the response targets. The functional role of inhibition was examined by conditionalizing firing rate on a high versus low rate in target neurons 90 ms to 30 ms before the saccade and examining the firing rate in both contralateral and ipsilateral neurons. Two models with racing diffusion processes were fit to the behavioral data, and the same analysis was performed on simulated paths in the diffusion processes that have been found to represent firing rate. The results produce converging evidence for the lack of a functional role for inhibition between neural populations corresponding to the two decisions.

2009 Special Issue: Neural mind reading of multi-dimensional decisions by monkey mid-brain activity

Brain-machine interfaces (BMIs) have the potential to improve the quality of life for individuals with disabilities. We engaged in the development of neural mind-reading techniques for cognitive BMIs to provide a readout of decision processes. We trained 2 monkeys on go/no-go tasks, and monitored the activity of groups of neurons in their mid-brain superior colliculus (SC). We designed a virtual decision function (VDF) reflecting the continuous progress of binary decisions on a single-trial basis, and applied it to the ensemble activity of SC neurons. Post hoc analyses using the VDF predicted the cue location as well as the monkeys motor choice (go or no-go) soon after the presentation of the cue. These results suggest that our neural mind-reading techniques have the potential to provide rapid real-time control of communication support devices.

Neural Prediction of Multidimensional Decisions in Monkey Superior Colliculus

SUMMARY To examine the function of the superior colliculus (SC) in decision-making processes and the application of its single trial activity for “neural mind reading,” we recorded from SC deep layers while two monkeys performed oculomotor go/no-go tasks. We have recently focused on monitoring single trial activities in single SC neurons, and designed a virtual decision function (VDF) to provide a good estimation of singledimensional decisions (go/no-go decisions for a cue presented at a specific visual field, a response field of each neuron). In this study, we used two VDFs for multidimensional decisions (go/no-go decisions at two cue locations) with the ensemble activity which was simultaneously recorded from a small group (4 to 6) of neurons at both sides of the SC. VDFs predicted cue locations as well as go/no-go decisions. These results suggest that monitoring of ensemble SC activity had sufficient capacity to predict multidimensional decisions on a trial-by-trial basis, which is an ideal candidate to serve for cognitive brain-machine interfaces (BMI) such as twodimensional word spellers.

Role of the anterior superior colliculus in rats

performance during training compared to wild-type (WT) mice when trained with 0.2 mA footshock. However, DIEDML mice learned the trace fear conditioning significantly faster and better compared to WT mice when trained using weaker training protocol (0.05 mA footshock). These observations suggest that DIEDML mice exhibit enhanced learning in trace fear conditioning and that up-regulation of CREB activity enhances even leaning as well as STM/LTM. We are now investigating the ability in learning of DIEDML mice using social recognition task. Research fund: KAKENHI 22022039.

Effect of bilateral lesion of superior colliculus on spontaneous movement in rats

The rat motor thalamic nuclei are composed of ventral medial (VM), ventral anterior (VA) and ventral lateral nuclei (VL). The caudodorsal portion of the VA–VL receives glutamatergic afferents from the cerebellum, whereas the VM and rostroventral portion of the VA–VL receive GABAergic afferrents from the basal ganglia. Previously we reported that axonal arborization was different between the rostroventral and caudodorsal VA–VL neurons by using single neuron-tracing method with Sindbis viral vector expressing membrane-targeted GFP. In the present study, the axonal arborization of single VM neurons was examined by the same method, and compared with the previous results of VA–VL neurons. When the axons exited from the thalamus, the reconstructed VM neurons always emitted axon collaterals to the thalamic reticular nucleus as VA–VL neurons. The VM neurons formed less axonal arborization in the striatum than rostrovental VA–VL neurons. In the cerebral cortex, the VM neurons sent axon fibers to more widespread cortical areas than VA–VL neurons, projecing to the primary motor, secondary motor, primary somatosensory, orbital, cingulate and insular areas. Of cortical layers, the axon fibers of VM neurons were most abundantly distributed in layer 1 (78.1 ± 5.6%), especially in the superficial part of layer 1. In comparison with the previously reported data of rostroventral (54%) and cuadodorsal VA–VL neurons (5.6%), VM neurons highly preferred layer 1 to other cortical layers. Although both the VM and rostroventral VA–VL have been reported to receive massive afferents mainly from the basal ganglia, the present results indicate that VM neurons more intensely innervate apical dendrites of pyramidal neurons in more widespread frontal/limbic areas than rostroventral VA–VL neurons. This suggests that, of motor thalamic neurons, VM neurons are most specialized to conrol the gain of widespread pyramidal neurons simultaneously.

Prediction of a Go/No-go Decision from Single-Trial Activities of Multiple Neurons in Monkey Superior Colliculus

The purpose of this study was to develop an algorithm capable of transforming neural activity to correctly report behavioral outcome during a cognitive task. We recorded from small groups of 2-5 neurons in the superior colliculus (SC) while monkeys performed a go/no-go task. Depending upon the color of a peripheral stimulus, the monkey was required to either make a saccade to the stimulus (go) or maintain fixation (no-go). In order to replicate the progress of the decision-making process and generate a virtual decision function ( VDF ), we performed a multiple regression analysis, with 1 msec resolution, on neuron activity during individual trials. Post hoc analyses by VDF predicted the monkeys choice with nearly 90% accuracy. These results suggest that monitoring of a limited number of SC neurons has sufficient capacity to predict go/no-go decisions on a trial-by-trial basis, and serves as an ideal candidate for a cognitive brain-machine interface (BMI).

Single-trial based neural prediction of immediate and delayed go/no-go decisions

Brain–computer interfaces (BCI) present a novel challenge to neuroscientists with their strict requirements for high reliability, real-time analysis and quantitative classification of multiple brain activity patterns. A BCI paradigm which is able to satisfy most of these requirements is the steady-state visual evoked potential (SSVEP) approach in which multiple flickering patterns evoke synchronized steady-state brain activity. In this study, we propose a multi-stage procedure for real-time BCI with an implementation for up to eight commands. Our EEG-based BCI system enables a user to navigate a small car on a screen in real time and to execute additional actions. This approach offers several novel points, such as integrated moving patterns for selective attention and minimal eye movement, as well as an online blind-source separation (BSS) unit for artifact rejection, improved feature selection and a fuzzy classifier. The modular and adaptive structure of the BCI platform allows an extension to an even higher number of commands, as well as to other BCI paradigms. O2P-K1Ø Single-trial based neural prediction of immediate and delayed go/no-go decisions Ryohei P. Hasegawa1,2, Yukako T. Hasegawa1,2, Mark A. Segraves2 1 Neuroscience Research Institute, AIST, Tsukuba, Japan; 2 Department of Neurobiology and Physiology, Northwestern University, USA