Brief analysis of procedure in optogenetics

This article aims to provide a brief analysis of the methods utilized in optogenetics data analysis procedure. The two main categories of procedure are online and offline analysis methods, which determines the causality of the system. Both approaches have their advantages and tradeoffs, and this article will introduce some examples of each procedure in an experimental setting, what are some specific processing methods they are using, what metrics are explained, how the two approaches performance compare, and how they produce and represent analytical data. One practical example on each approach, performed on a dataset from a previous experiment imaging neuronal activity in rats, is provided for better clarity and comparison measures. The offline example fits gaussian mixture model on high-pass filtered data, and the separate gaussian models are used to predict the status of the neuron given its relative fluorescence intensity. The online example uses an existing algorithm called OASIS, which uses an autoregressive model to reconstruct calcium trace and thus infers the spike. The main difference between the two is demonstrated to be their robustness and accuracy: online approach is more robust and can be utilized while recording the data, giving interpretable results with low latency, yet its accuracy does not depend on obtained sample number; offline approach is more time-consuming while fitting and training data to an optimal model. However, offline approaches accuracy will increase with large sample size. Both approaches provide deep insight into the acquired datasets, and while analyzing data they should be used strategically to fit the specific needs of the task.