Motor cortical networks for skilled movements have dynamic properties that are related to accurate reaching

Roles of Motor Cortex Neuron Classes in Reach-Related Modulation for Hemiparkinsonian Rats

Disruption of the function of the primary motor cortex (M1) is thought to play a critical role in motor dysfunction in Parkinson's disease (PD). Detailed information regarding the specific aspects of M1 circuits that become abnormal is lacking. We recorded single units and local field potentials (LFPs) of M1 neurons in unilateral 6-hydroxydopamine (6-OHDA) lesion rats and control rats to assess the impact of dopamine (DA) cell loss during rest and a forelimb reaching task. Our results indicated that M1 neurons can be classified into two groups (putative pyramidal neurons and putative interneurons) and that 6-OHDA could modify the activity of different M1 subpopulations to a large extent. Reduced activation of putative pyramidal neurons during inattentive rest and reaching was observed. In addition, 6-OHDA intoxication was associated with an increase in certain LFP frequencies, especially those in the beta range (broadly defined here as any frequency between 12 and 35 Hz), which become pathologically exaggerated throughout cortico-basal ganglia circuits after dopamine depletion. Furthermore, assessment of different spike-LFP coupling parameters revealed that the putative pyramidal neurons were particularly prone to being phase-locked to ongoing cortical oscillations at 12-35 Hz during reaching. Conversely, putative interneurons were neither hypoactive nor synchronized to ongoing cortical oscillations. These data collectively demonstrate a neuron type-selective alteration in the M1 in hemiparkinsonian rats. These alterations hamper the ability of the M1 to contribute to motor conduction and are likely some of the main contributors to motor impairments in PD.

Investigation into machine learning algorithms as applied to motor cortex signals for classification of movement stages

Neuroinformatics has recently emerged as a powerful field for the statistical analysis of neural data. This study uses machine learning techniques to analyze neural spiking activities within a population of neurons with the aim of finding spiking patterns associated with different stages of movement. Neural data was recorded during many experimental trials of a cat performing a skilled reach and withdrawal task. Using Weka and the LibSVM classifier, movement stages of the skilled task were identified with a high degree of certainty achieving an area-under-curve (AUC) of the Receiver Operating Characteristic of between 0.900 and 0.997 for the combined data set. Through feature selection, the identification of significant neurons has been made easier. Given this encouraging classification performance, the extension to automatic classification and updating of control models for use with neural prostheses will enable regular adjustments capable of compensating for neural changes.