Latency Detection in Motor Responses: A Model-Based Approach With Genetic Algorithm Optimization

The latency of a response is one of the most frequently reported parameters when describing the characteristics of a motor system. Such measurement provides important information both to the basic researcher investigating the neural circuitry of the underlying physiological system and to the clinician gathering information for diagnosing a patient. Our concern here is that when the latency of a response is determined on experimentally recorded data by using the most commonly referenced techniques to find the onset of a motor response, the resulting figure encompasses both the neural processing time and the dynamics of the system producing the response (e.g., the musculoskeletal apparatus). Therefore, the resulting latency measurement cumulates information relative to two substantially different sources and thus having different implications. The goal of our study is that of suggesting a technique allowing the separation of the relative contributions of neural transmission and processing time from that of the dynamics of the motor system. This is accomplished by using a technique based on fitting a model to the experimentally recorded response, thus allowing to exploit as much as is known with regards to the dynamics of the studied motor system (e.g., model order and constraints on the values of the model parameters). The optimization of the model parameters for fitting the experimental data is carried out using a real-valued genetic algorithm, allowing to avoid trapping in local, suboptimal minima. The use of this approach allows to estimate the pure delay in the response introduced by neural processing more accurately than the traditional latency detection techniques based on adaptive thresholds