Rolf Eckmiller

An on-line spike form discriminator for extracellular recordings based on an analog correlation technique

The discrimination of single unit activity in extracellular recordings presents a serious problem when the signal-to-noise ratio is low or when the amplitudes of interspersed spikes are similar. By exploiting spike form, the system described here performs discrimination using on-line hardware template matching. Using analog delay lines, the combined deviation of 8 input signal values from 8 stored template values is calculated simultaneously. The 8 template values are selected by adjusting 8 cursors to the desired spike trace on a CRT; the spike form discriminator (SPIFODIS) then generates a deviation function which steeply drops to zero whenever form similarity occurs, allowing for easy triggering. The performance of SPIFODIS was compared quantitatively with that of a conventional amplitude trigger in two cases: when detecting a single unit with varied signal-to-noise ratios and when separating double units of equal amplitude. At signal-to-noise ratios between 2 and 1 the error rate for SPIFODIS was only 15-50% of that of an amplitude trigger. In double-unit recordings showing only form differences, spikes are discriminated with very low error rate, while an amplitude trigger fails completely.

Video-oculographic measurement of 3-dimensional eye rotations.

An effective and relatively calibration-free method for the measurement of 3-dimensional eye rotations is described. The experimental method consists of recording with a video camera the positions of distinct globe markers provided by a tight-fitting, soft contact lens before and after an eye rotation. The mathematical analysis represents the eye rotation by a rotation vector (rotation axis and angle) and three, spherical Euler angles. The method is linear over the entire range of natural eye rotation, the resolution limited only by the video set-up used. Due to its easy applicability, the system is well suited for various eye movement measurements in clinical or research environments.

Afferent innervation of extraocular muscles in the rat studied by retrograde and anterograde horseradish peroxidase transport

After injection of horseradish peroxidase (HRP) into extraocular muscles of rat perikarya were labeled mainly along the medial edge of the ophthalmic subdivision of the trigeminal ganglion but not in the mesencephalic nucleus of the trigeminal nerve. Injections of HRP into the trigeminal ganglion labeled simple as well as branching and meandering free fiber endings in extraocular muscles. No evidence for muscle spindles was found, but the meandering endings may be considered as candidates for stretch receptors.

Neural control of foveal pursuit versus saccadic eye movements in primates — Single-unit data and models

A framework for two oculomotor subsystems, an electromechanical pursuit system and the saccadic system, is proposed on the basis of neurophysiological findings in alert monkeys. This model is arranged into a sequence of spatiotemporal translation, motor program generation, and neural integration. For the purpose of foveal pursuit movements, an eye velocity signal must be internally generated and updated according to the current position error on the retina. This design probably involves a recently discovered class of pursuit neurons. Saccadic eye movements are probably controlled by burst neurons. Based on striking similarities between the neural activity of these recently analyzed burst neurons and Renshaw cells, a detailed quantitative model for the generation of preprogrammed saccades, the Renshaw model, is proposed. Open and controversial questions in the literature concerning various strategies being used by the primate oculomotor system are mentioned.

Control of arm movements in a 2-dimensional pointing task

The present study analyses in humans the control principles of sequential, unpracticed pointing movements in a 2-dimensional space. Our data reveal that variable pointing errors add up within such sequences. This finding supports the hypothesis that movement amplitude rather than position is the controlled variable of the investigated movements.

Interaction of visual and non-visual signals in the initiation of smooth pursuit eye movements in primates

The initiation of smooth pursuit eye movements (PEM) by visual and non-visual signals was analysed in humans and monkeys. While PEM latency ranged around 150 ms when a purely visual target was provided, it often dropped to about 0 ms, or even became negative, when target movement was coupled to the subjects arm; this suggests that signals about the intention to move the arm can be evaluated for PEM control. Eye movements always started in the visually correct direction, independent of the sign of coupling between arm and target; from this we conclude that intentional signals are not mere triggers, but also convey directional information. Short-latency PEM trials were intermixed with those characterized by normal latencies, which often resulted in bimodal latency distributions; this suggests that visual and intentional signals compete for the control of PEM.

Computational model of the motor program generator for pursuit

A parallel processing neural network model of a motor program generator (MPG) for pursuit eye movements (PEM) was developed. The MPG model consists of two neural networks (velocity maps), which represent velocity values theta R and theta L respectively, as eccentric locations on the map with zero in the center. Neurons are arranged in a circular layer and connected only to their immediate neighbors. The potential field P of all neurons is analogous to a flat circular membrane whose center can be pushed up or down. During PEM one of the two maps, which are connected in a push-pull fashion, always features an activity peak (AP) which travels with constant velocity vT from one neuron to the next. The shape of P defines whether AP travels in a circle (theta = constant), towards the periphery (theta increase) or towards the center (theta decrease). Such a model provides a novel approach for understanding neural generators of non-periodical motor programs.

Dynamic adaptation of the blind pointing characteristic to stepwise lateral tilts of body, head, and trunk

Blind pointing (i.e. pointing to visual targets without seeing the pointing arm) was investigated in normal subjects in response to stepwise lateral tilts (20 degrees) of the body, head, and trunk. Blind pointing positions of the right index finger on the outer surface of a hemispherical screen were measured relative to the positions of visual targets that were presented along a horizontal line (+/- 30 degrees in head coordinates) on the inner screen surface, thus yielding a blind pointing characteristic (BPC). (1) BPC is highly reproducible and can be subdivided into separate branches for the ipsi- and contralateral hemifields. These branches are rotated relative to the target line by individually different BPC angles pi i and pi c. (2) pi i exhibits characteristic time courses (measured within 10 min following a stepwise tilt) for each paradigm. (3) Body tilt (left ear down) causes a step-like increase in pi i of up to 14 degrees; body tilt (right ear down) causes a step-like decrease in pi i to about zero. (4) Trunk tilt (right shoulder down) produces a gradual decrease in pi i of up to 6 degrees (average time constant tau T = 5 min); trunk tilt (left shoulder down) produces a gradual increase in pi i of up to 4 degrees. (5) Head tilt (left ear down) causes an increase in pi i of up to 9 degrees followed by a gradual decrease (average time constant tau H = 6 min); head tilt (right ear down) causes a step-like decrease in pi i with unsignificant further changes. These findings are discussed in terms of a neural sensorimotor coordinate transformation process receiving separate, dynamic otolith and neck afferent influences.

A general numerical method evaluating three-dimensional eye rotations by scanning laser ophthalmoscopy.

A general computational method is described to specify completely the rotational state of the eye in three dimensions by scanning laser ophthalmoscopy (SLO). The method uses the simplex algorithm to fit the eyes rotational parameters to data given by n individually selected ocular fundus landmarks before and after the eye rotation.

Neural Computers for Motor Control

The complexity of biological motor control is first demonstrated with the surprising lack of precision in 2-dimensional blind pointing movements (without vision of the pointing arm) in human subjects and with several deficits in the blind pointing performance of trained monkeys following a unilateral lesion of the parietal cortex. Subsequently, several novel approaches to designing neural networks for motor control of a redundant robot arm for 2-dimensional movements are described. A comparison of these biological and technical motor systems under neural net control clearly demonstrates the importance of an internal representation of space with continuous updating from various sensors regarding targets, obstacles, and the controlled arm in space.