Simulation of the behavior of fine and gross motor skills of an individual with motor disabilities

We have developed a neural network model that imitates the central nervous system’s control of motor sensors (Sánchez-Torres and Rodríguez-Romo in Neurocomputing 581:127511, 2024). Our research explored various levels of connectivity in our neural network related to neuroplasticity in the central nervous system. We have conducted a study comparing healthy individuals to those with motor impairments by utilizing reinforcement learning and transfer entropy. In our previous research (Sánchez-Torres and Rodríguez-Romo in Neurocomputing 581:127511, 2024), we have simulated human walking while encountering obstacles as an instance of gross motor activities. Now, we have used the same model to simulate fine motor activities. Our goal is to identify differences in information transmission between gross and fine motor activities among healthy individuals and those with motor impairments by evaluating the effective connectivity of our network. To regulate learning accuracy in our model, we introduced a variable called numClusterToFire. However, we discovered that the value for this variable requires careful calibration. If the value is too small, agent exploration is insufficient, and network learning is inefficient. Conversely, learning times increase exponentially, often unnecessarily if the value is too large. We conducted simulations for gross and fine motor skills using three different numClusterToFire values and found that as we increased numClusterToFire, the time required for the network to memorize the outputs for each of the objects in the test set also increased. Our findings indicate that in gross motor skills, which do not require precision, changes in the numClusterToFire variable do not affect information transfer behavior. Conversely, in fine motor skills, information transfer decreases as numClusterToFire increases. On the other hand, our model revealed that for healthy and disabled individuals, the transfer of information between the input layer and the first hidden layer is higher for fine motor skills; this important biological fact suggests the influence of external cues in performing this activity successfully. Additionally, our neural network model showed that movements that do not require precision do not necessarily require a high level of neuroplasticity. Increasing neuroplasticity may cause some neurons to transmit more information than others. Whereas, increasing neuroplasticity through practice is essential for precise movements like fine motor skills. We also found that information transfer in the network’s hidden layers is similar for fine and gross motor activities, as we observed identical patterns. However, the distribution and proportion of these patterns differ, concluding that more neurons are involved in fine motor activities, and more information is transferred compared to gross motor activities. Finally, a pattern was observed in the transfer of information in the last hidden layer, which is only present in fine motor skills. This pattern is associated with the precision of the movements.