Implementing (simplified) MOD method for detecting Miniature events

[schloegl]

Detecting of Miniature events is an import aspect. Currently, Template matching and Deconvolution methods are implemented in
Stimfit. The "machine-learning approach optimal-detection" [1] can provide better detection performance. The aim is to implement a (simplied) version of the MOD method for interactive use within Stimfit. This requires the following tasks

• annotating events (including importing and exporting to some annotation file
• computing the optimal filter
• computing the ROC and AUC curve
• computing optimal threshold based on Cohen's kappa
• applying the method to all (scored and unscored) data, producing the same output as Template Matching and deconvolution approach.

[1] Xiaomin Zhang, Alois Schlögl, David Vandael, and Peter Jonas
MOD: A novel machine-learning optimal-filtering method for accurate and efficient detection of subthreshold synaptic events in vivo.
Journal of Neuroscience Methods, Volume 357, 1 June 2021, 109125.
https://www.sciencedirect.com/science/article/pii/S0165027021000601
https://doi.org/10.1016/j.jneumeth.2021.109125