Predicting the Temporal Dynamics of Prosthetic Vision

Code for "Predicting the Temporal Dynamics of Prosthetic Vision" at IEEE EMBC 2024.

Please cite as:

Y Hou, L Pullela, J Su, S Aluru, S Sista, X Lu, M Beyeler (2024). Predicting the Temporal Dynamics of Prosthetic Vision. 46th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (IEEE EMBC)

Note: Y Hou and L Pullela contribute equally to this work.

In this paper, we introduced two computational models designed to accurately predict phosphene fading and persistence under varying stimulus conditions, cross-validated on behavioral data reported by nine users of the Argus II Retinal Prosthesis System.

Setup

To run the notebooks, please install the required Python packages via pip:

pip install -r requirements.txt

File description

• baseline_training.ipynb and baseline_evaluation.ipynb: baseline model training and evalution. Code adapted from Avraham, David, et al. “Retinal Prosthetic Vision Simulation: Temporal Aspects.” Journal of Neural Engineering, vol. 18, no. 4, IOP Publishing, Aug. 2021, p. 0460d9, doi:10.1088/1741-2552/ac1b6c.
• exponential_cross_subject.ipynb: Exponential model trained and evaluted using leave-one-subject-out cross-validation.
• exponential_cross_stimulus.ipynb: Exponential model trained and evaluted using leave-one-stimulus-condition-out cross-validation.
• FFT_cross_subject.ipynb: FFT model trained and evaluted using leave-one-subject-out cross-validation.
• FFT_cross_stimulus.ipynb: FFT model trained and evaluted using leave-one-stimulus-condition-out cross-validation.