Modelling of axonal fields in the optic nerve for direct MR detection studies

Abstract

A number of studies have now shown that direct detection of neuronal firing by MRI may be possible. The optic nerve carries all visual information from the eye to the brain and is a particularly promising target for these measurements. However, it has been assumed that the effects of axonal firing may not be detectable, as a single firing event produces a bipolar waveform of around 1 ms duration whose effects should cancel on MR. A simulation of the magnetic modulation which could feasibly be produced by the optic nerve over an extended period and with different firing rates has been developed. The Hodgkin–Huxley equations were calculated for an array of model ganglion cell axons which were assumed to act as voltage to pulse frequency converters. Dependence of the modulating waveform on relative action potential firing start time was investigated. Although the simulated waveforms were bipolar at the beginning, during the period of MR acquisition the different frequencies combine to produce a largely positive waveform. The simulation included only contrast luminance changes and not color, spatial correlations or other more sophisticated processing in the retina. A gradient echo sequence was used at 3 T to create images for analysis by the ghost reconstructed alternating current estimation (GRACE) method from phantoms subject to current modulation by the actual modelled axonal waveforms. The optic nerve was also imaged using the same method during visual stimulus by a strobe light in adult human volunteers at 3 T. Analysis of digitized video recordings of eye locations during strobe stimulation outside the magnet showed no correlation with the applied strobe frequency over the short duration of the scans. Images of the optic nerves at an echo time of TE = 39 ms had weak but significant first harmonic ghosts in locations consistent with the applied stimulus as calculated from GRACE theory in just two out of thirteen studies on 10 volunteers and a detection rate of only 15% providing no clear evidence for direct detection in these experiments.