5.2. Materials and methods 5 65 detection threshold (explained in sectionSection 5.2.4), the phosphene is activated with a brightness equal to the sigmoidal activation 1 1+e−λ(A−A50) , (5.10) where λis the slope of the sigmoidal curve and A50 is the value of Afor which the phosphene reaches half the maximum brightness. 5.2.4. Stimulation threshold Our simulator uses a thresholding model based on psychometric data from (Fernández et al., 2021). Phosphenes are only generated when the cortical tissue activation (explained in section Section 5.2.3) reaches the activation threshold Athr, which is obtained for each electrode separately upon initialization of the simulator. To introduce a degree of variability between electrodes, Athr is sampled from the normal distribution N(Th50,σ 2). (5.11) The default values of the 50% probability thresholdTh50, and the standard deviation σare fit on data from (Fernández et al., 2021) and can be found in sectionSection 5.2.6. Note that, by default, the detection thresholds remain constant after initialization. However, in accordance to the user requirements, the values can be flexibly adjusted or re-initialized manually. 5.2.5. Temporal dynamics Using a memory trace of the stimulation history, the simulator accounts for basic accommodation effects on brightness for prolonged or repeated stimulation, as described in prior work (Schmidt et al., 1996). Each frame, the memory trace is dynamically updated as follows: Qt =Qt−∆t +∆Q (5.12) with ∆Q=µ− Qt−∆t τtrace + Ieff ·κ¶∆t. (5.13) Here, τtrace is the time constant of the trace decay in seconds, and the parameter κ controls the input effect. Note that the memory trace is used for the phosphene brightness and not the phosphene size. Because there is little experimental data on the temporal dynamics of phosphene size in relation to the accumulated charge, only the instantaneous current is used in the calculation of the phosphene size. 5.2.6. Parameter estimates By default, our model uses the parameters specified below. Unless stated otherwise, these parameter estimates were obtained by fitting our model to experimental data using the SciPy Python package, version 1.9.0 (Virtanen et al., 2020). More details on the comparison between the models’ estimates and the experimental data can be found in the next section. Note that the parameter settings may strongly depend on the specific experimental conditions (such as the type of electrodes).
RkJQdWJsaXNoZXIy MTk4NDMw