README.md

---
license: pddl
tags:
- eeg
- medical
- clinical
- classification
- parkinson
- interval estimation
---
# Singh2021: EEG Parkinson's Classification Dataset with Interval Estimation Task
The Singh2021 dataset contains EEG recordings collected during an interval timing task designed to study cognitive control in individuals with Parkinson's disease (PD). The dataset contains a total of 83 PD patients and 37 demographically matched healthy controls. Most PD patients (n = 74) completed the task while ON medication, and a subset (n = 9) completed both ON and OFF dopaminergic medication sessions.

Participants performed a peak-interval timing task with intermixed 3-second and 7-second trials. They were instructed to press a key when they estimated the target interval had elapsed. Visual distractions were included to discourage counting. Each participant completed 80 trials (40 per interval type). Only trials with a minimum of 20 valid keypresses per interval condition were included in analyses.

EEG was recorded using a 64-channel actiCAP system at 500 Hz.
## Paper
Singh, A., Cole, R. C., Espinoza, A. I., Evans, A., Cao, S., Cavanagh, J. F., & Narayanan, N. S. (2021). **Timing variability and midfrontal~ 4 Hz rhythms correlate with cognition in Parkinson’s disease**. _npj Parkinson's Disease_, 7(1), 14.

DISCLAIMER: We (DISCO) are NOT the owners or creators of this dataset, but we merely uploaded it here, to support our's ([EEG-Bench](https://github.com/ETH-DISCO/EEG-Bench)) and other's work on EEG benchmarking.
## Dataset Structure
- `data/` contains the raw experiment EEG data.
- `Copy_of_IntervalTiming_Subj_Info_AIE.xlsx` contains information about the participants (except the 9 ON+OFF PD participants).

Note that 6 PD (`1205`, `1255`, `1345`, `1355`, `1495` and `1545`) and 4 control (`1005`, `1045`, `1165` and `1355`) patients' recordings were excluded due to artifacts and noise.
### Filename Format
A recording session consists of 3 files:
- `[GROUP][SID].vhdr`: The header file with meta information
- `[GROUP][SID].eeg`: contains the EEG (and accelerometer) data
- `[GROUP][SID].vmrk`: contains event information

where GROUP is either `Control` or `PD` (Parkinson's Disease) and SID is the session ID.

While the paper gives some averaged information like age and MOCA scores about the 9 ON+OFF PD participants, the corresponding recordings must be inferred to be those 18 missing from `Copy_of_IntervalTiming_Subj_Info_AIE.xlsx`. If it is desired to identify the 9 pairs of recordings belonging to each of the 9 ON+OFF PD participants, one can make an educated guess from similarities in the session ID namings:
- `1815` (ON) and `2865` (OFF)
- `1835` (ON) and `2835` (OFF)
- `1845` (ON) and `2845` (OFF)
- `1855` (ON) and `2855` (OFF)
- `1865` (ON) and `2865` (OFF)
- `3445` (ON) and `2445` (OFF)
- `3515` (ON) and `2515` (OFF)
- `3565` (ON) and `2565` (OFF)
- `3625` (ON) and `2625` (OFF)

where ON refers to the session ON medication and OFF to the session OFF medication (withdrawal 12h prior to recording).

To distinguish ON and OFF recordings, we assumed that ON sessions were recorded prior to OFF sessions (as mentioned in the paper: "We first tested [...] (ON sessions), and then [...] (OFF sessions)", p. 14: Methods->Paricipants) and used the measurement time information stored in each recording to determine the earlier session as the ON session.

To further support the hypothesis that these session pairs do belong to the same participant, one may notice that the measurement times of each pair is always precisely one day apart (allowing for the 12h window to withdraw from medication).

### Fields in each File
In python, the 3 files that make up a raw recording can be read via:
```python
import mne
raw = mne.io.read_raw_brainvision("path_to/[GROUP][SID].vhdr")
```
Now, `raw.get_data(units='uV')` yields a numpy array of shape `(#channels, time_len)` in micro-Volt units.

Some general info can be inspected with `raw.info`, such as the sampling rate (`raw.info["sfreq"]`), as well as the measurement time (`raw.info["meas_date"]`).

The channel names (in their correct order) can be seen via `raw.ch_names`.

Events can be read with
```python
events_list, events_dict = mne.events_from_annotations(raw)
```
where `events_dict` contains the mapping of the original event types (like `"Stimulus/S  1"`) to event IDs in `[1,2,...,7,255]`, the latter of which are used in `events_list`.

`events_list` is a list of events, ordered by time. Each entry `e = [timestamp, (not important), event ID]` consists of the time of the event onset `timestamp` that refers to the `time_len` dimension in the `raw.get_data()` EEG array, as well as the event-ID.

(See the `https://mne.tools/stable/generated/mne.io.Raw.html` documentation for more details.)

The different event types will likely have the following meaning (inferred by `events_list` inspection via trial description from the paper):
- `Stimulus/S  1`: Instruction to estimate short (3s) interval is shown on screen for 1s.
- `Stimulus/S  2`: Instruction to estimate long (7s) interval is shown on screen for 1s.
- `Stimulus/S  3`: Blue rectangle "GO" cue is shown on screen.
- `Stimulus/S  4`: Participant presses `spacebar` key.
- `Stimulus/S  5`: Participant releases `spacebar` key.
- `Stimulus/S  6`: Distracting vowel appears on screen.
- `Stimulus/S  7`: Trial Feedback is shown on screen.
- `Stimulus/S255`: End of last trial

## License
By the original authors of this work, this work has been licensed under the PDDL v1.0 license (see LICENSE.txt).