logLIRA: a method for reliable suppression of electrical stimulation artifacts enabling short-latency neural response recovery

This dataset provides a semisynthetic benchmark for evaluating methods that detect and remove electrical stimulation artifacts in neural recordings, often referred to as Stimulus Artifact Rejection (SAR) algorithms. The dataset is organized into a ZIP archive containing the following components: lookup_table.xlsx: an Excel file mapping each item of the dataset to its corresponding data source, enabling traceability and reproducibility. basal: a folder containing 2.5-minute chunks of basal neural activity. These segments are randomly selected to meet a target mean firing rate (MFR) and are checked to be free of recording artifacts (e.g., movement). templates: a library of clustered and averaged stimulation artifacts extracted from single channels belonging to recordings with Activity-Dependent Stimulation (ADS). These templates serve as the basis for generating semisynthetic stimulation artifacts in the benchmarking snippets. snippets: semisynthetic neural activity snippets that include stimulation artifacts. These are generated by superimposing artifact templates onto basal activity, simulating realistic conditions for stimulation artifact removal.