Paradoxical lesions, plasticity and active inference

Abstract

Paradoxical lesions are secondary brain lesions that ameliorate functional deficits caused by the initial insult. This effect has been explained in several ways; particularly by the reduction of functional inhibition, or by increases in the excitatory-to-inhibitory synaptic balance within perilesional tissue. In this article, we simulate how and when a modification of the excitatory-inhibitory balance triggers the reversal of a functional deficit caused by a primary lesion. For this, we introduce in-silico lesions to an active inference model of auditory word repetition. The first in-silico lesion simulated damage to the extrinsic (between regions) connectivity causing a functional deficit that did not fully resolve over 100 trials of a word repetition task. The second lesion was implemented in the intrinsic (within region) connectivity, compromising the model's ability to rebalance excitatory-inhibitory connections during learning. We found that when the second lesion was mild, there was an increase in experience-dependent plasticity that enhanced performance relative to a single lesion. This paradoxical lesion effect disappeared when the second lesion was more severe because plasticity-related changes were disproportionately amplified in the intrinsic connectivity, relative to lesioned extrinsic connections. Finally, this framework was used to predict the physiological correlates of paradoxical lesions. This formal approach provides new insights into the computational and neurophysiological mechanisms that allow some patients to recover after large or multiple lesions.

Keywords: active inference; learning; paradoxical lesions; plasticity; structure–function relationship.

Graphical Abstract

Figure 1

Proportion correct. The line plot shows the stimulated (mean) proportion of correct responses for each model across the 100 trials, with 95% confidence interval. The x-axis is the trial number and the y-axis is the correct number of responses (%). Blue line reports the control model, black line reports the model with single lesion, green line reports the model with dual lesion no. 1 (severe dual lesion) and magenta line reports the model with dual lesion no. 2 (mild dual lesion). The vertical black dashed line represents when the first lesion was introduced (1st trial), and the vertical grey dashed line represents when the second lesion was introduced (20th trial) to the model.

Figure 2

Measuring plasticity in extrinsic (likelihood) connections A— the first lesion location. (A) Plots plasticity in the first lesion location, for each model across 100 trials, with 95% confidence intervals and (B) shows the gradients (rate of change) for the plasticity-related changes for the lesioned models over 10 trials to ensure smoothing out of any noise. (A) The x-axis is the trial number and the y-axis represents the KL-divergence (measured in nats) between initial and current distribution. Blue line reports the control model, black line reports the model with single lesion, green line reports dual lesion model no. 1 (with the severe dual lesion), magenta line reports dual lesion model no. 2 (with the mild dual lesion). The vertical black dashed line indicates when the first lesion was introduced (1st trial), and the vertical grey dashed line indicates when the second lesion was introduced (20th trial). (B) The x-axis is the trial number and the y-axis represents the gradient for all lesioned models.

Figure 3

Measuring plasticity in intrinsic (prior transition) connections B—the second lesion location. This figure uses the same format as previous figure: (A) plots plasticity in the second lesion location, for each model across 100 trials, with 95% confidence intervals and (B) shows the gradients (rate of change) for the plasticity-related changes for the lesioned models. (A) The x-axis is the trial number and the y-axis represents the KL-divergence (measured in nats) between initial and current distribution. (B) The x-axis is the trial number and the y-axis represents the gradient for the dual lesioned models.

Figure 4

Simulated local field potentials. These plots show the simulated local field potentials for each model for a target word across the 100 trials (x-axis). These are plotted after bandpass filtering between 4 and 32 Hz (Friston et al., 2017a). This is calculated from membrane depolarization (i.e. post-synaptic potential) gradients computed using the inputs from other neurons. The blue shows the trajectory of evoked responses over arbitrary units (y-axis), where positive indicates excitatory responses and negative indicates inhibitory responses. The top row presents the control model, the second row shows the single lesion model and the last two rows show the dual lesion models. Each plot represents a single instantiation of the simulated models.