What was new in previous releases?

Version 0.2

Notable changes

• Added the HilbertDetector and optimized its performance on long recordings

Authors

• Adam Li

• Patrick Myers

Detailed list of changes

This release is the first one under the mne.tools umbrella. We introduced three core detectors that we found were used and cited in the literature: Line Length, RMS and Hilbert detectors. We have organized a roadmap for what future improvements would entail. In addition, we have added tutorials that are rendered as jupyter notebooks, which walk through usage of the package to: i) load data from BIDS, ii) run a detection and then iii) evaluate the efficacy using tools from scikit-learn.

Enhancements

• Added mne_hfo.compute_chs_hfo_rates() to compute HFO rates per unit of time for every channel, by Adam Li (#13)

• Added mne_hfo.io.create_events_df to generate a DataFrame of HFO events from Raw object, or dictionary of lists of HFO endpoints, by Adam Li (#7)

• Added mne_hfo.find_coincident_events() to compare two dicts that contain event information by Patrick Myers (#10)

• Added notebook to demo use of detection algorithms by Patrick Myers (#10)

• Vectorized detection overlap check to enhance scoring speed by Patrick Myers (#15)

• Added notebook to demo use of GridSearchCV to optimize detector performance by Patrick Myers (#15)

• Added module to compare detections and notebook to demo usage by Patrick Myers (#22)

• Added initial implementation of HilbertDetector by Patrick Myers (#23)

• Improve memory utilization by allowing parallelization of the entire workflow per channel by Patrick Myers (#38)

API changes

• Added mne_hfo.io.events_to_annotations to go from *events.tsv to *annotations.tsv files, by Adam Li (#10)

• Added mne_hfo.sklearn.make_Xy_sklearn() to format data into scikit-learn compatible data structures for the sake of running hyper-parameter searches with SearchCV functions, by Adam Li (#15)

• Separated postprocessing step into two discrete steps _threshold_statistic and _post_process_ch_hfos by Patrick Myers (#23)

Requirements

• Updated requirement version for mne to v0.23+, by Adam Li (#44)

• Added tqdm, joblib and pandas to requirements, by Adam Li (#7)

Bug fixes

• Fixed channel name issue introduced by redundant type checks when using fit_and_predict by Patrick Myers (#15)