Work Description

Date: 10/24/22

Dataset Title: Epileptic High Frequency Oscillation EEG with Network Analysis

Dataset Creators: Lin J & Stacey W

Dataset Contact: William Stacey, M.D., Ph.D. wstacey@umich.edu

Code repositiory for data processing scripts
https://github.com/J4KLin/HFO-Network (preservation versions as of Nov. 22, 2022 also accompany the dataset deposited into Deep Blue Data).

Reserach Overview:
High frequency oscillation (HFO) has been known as a promising electrographic biomarker for epiletic tissue for decades. To this end, we characterized HFO networks through functional connectivity analysis of clinical intracranial EEG data from patients who have undergone resective surgery (pre-processed EEG dataset provided for one good outcome Engel I [UMHS-0028] and one poor outcome Engel III [UMHS-0030] patient). From the networks, we performed centrality analyses (results of which are also provided) to evaluate how HFO features can be used to predict patient outcome.

Data Flow: (see 'Details on file contents' below for more information)
Raw EEG during HFO and nonHFO events are preprocessed through 80-500 Hz filtering and smoothed (root mean square). For each patient, the HFO rate along with the functional connectivity networks are first characterized with their centralities computed. For the HFO rates and centralities, the critical resection percentage (CReP) and centrality ranks grouped into different percentiles for each electrode type are extracted.

Software requirements:
Matlab R2021a Version 9.10 or above (no other toolboxes necessary)

Instructions:
1. Unzip "Data.zip" file.
/BKGEEG - Preprocessed background (nonHFO) EEG data (.mat)
/CombinedData - Final centrality and hfo features for all patients (.mat)
/HFOEEG - Preprocessed HFO specific EEG data (.mat)
/PatientData - Patient level data (.mat)

2. Download and move all "UMHS-xxxx-elecxx.mat" files into the /HFOEEG folder.

3. Download up-to-date code from GitHub repository (above) and move to same directory as the /Data folder to run.
For a given patient, change the patient identifier number under %% CHANGE ME %%.

Details on file contents:
/Data/BKGEEG/UMHS-xxxx.mat
- Background EEG data that is filtered (80-500 Hz) and smoothed (root mean square RMS)

/Data/HFOEEG/UMHS-xxxx-elecXX.mat
- Up to 500 samples of EEG data during which a HFO was detected on the specified electrode. Like background data this is also filtred and smoothed.

/Data/CombinedData/combinedPatientData.mat
- All centrality and HFO rate features for all patients. Includes critical resection percentage (CReP) and centrality ranks grouped into different percentiles for each electrode type.

/Data/PatientData/UMHS-xxxx.mat
- Patient level data including HFO information, electrode information, patient surgical outcome, and centrality. The outputs of 'getCReP.m' and 'getPercentileRanks.m' for CReP information and centrality ranks grouped into different percentiles are respectively saved under the variables 'crepTable' and 'prtileTable.'

Variable Abbreviations and Disambiguations:
SOZ: Seizure onset zone electrodes
RV: Resected volume electrodes
BOTH: SOZ & RV
REST: Neither SOZ nor RV
perSOZinRV (SOZRVpart): Percentage of SOZ that is in the RV
[-High] subclass: SOZRVpart >= .8
[-Low] subclass: SOZRVpart < .8

Related publication(s):
Lin J, Zochowski M, Shedden K, Stacey W,
"Network dynamics of interictal High Frequency Oscillations predict surgical outcome within the clinical workflow in refractory epilepsy,"
Undergoing submission