...
Each channel is saved as a MATLAB timeseries object, with asynchronous (that is, not necessarily uniformly sampled) time values.
File | Structure | Fields | Subfields | Description | ||
---|---|---|---|---|---|---|
phys[protocol_id].mat | phys* | .study_id | link to study in database; 0 before saving * saved as "-struct", so individual variables loaded with phys = load(...) | |||
.name | name | |||||
.start_time (optional) | date-time: precise start time of recording
Optional because for some fMRI physiology we don't have this, or it's not synced with scanner. | |||||
.num_chans | ||||||
.chan{1:N} | cell array of channels | |||||
.name | channel name | |||||
.type | "rec" (recorded) or "calc" (calculated) | |||||
.filename | mat file of timeseries object; variable has channel name. File is located in phys folder). The channels are synchronized such that the time is equivalent (this is critical for events). The time should start at 0. Time is in seconds. | |||||
.ts (memory only) | the timeseries when loaded; this is not saved in the file, and is only loaded when needed. It slows things down a lot if it gets loaded every time. | |||||
.sample_rate | empty for asynchronous | |||||
.offset | time offset (relative to phys.start_time if present) | |||||
.calc_sourcechans | for calculated channels; cell array of "rec" channel names used to calculate "calc" channel | |||||
.descr | description (optional) | |||||
.units | units (optional) | |||||
.num_features | number of features (types of markers, eg. resp peaks/troughs, ECG Q/R/S | |||||
.feature{1:N} | cell array of features (optional) | |||||
.type | feature type (peak/trough, ECG) | |||||
.descr | description (optional) | |||||
.num_markers | ||||||
.marker_list | ||||||
.marker | cell array or markers | |||||
.name | ||||||
.x | ||||||
.y | ||||||
.shortname | ||||||
.color | ||||||
.num_events | this includes states (events with non-zero duration) | |||||
.event{1:N} | Cell array of events or states; note that these are all events, not only those in the database. For analysis, we will use the database events and select segments of the timeseries. For this reason, the names don't matter too much, as we can select events for analysis and classify them at the same time we incorporate them into the database. | |||||
.time | time in seconds, relative to phys.start_time | |||||
.event_id | event_id from database | |||||
.duration | empty or 0 for event, time for "state" | |||||
.type (optional?) |
Example (phys file incorrect size - earlier version where timeseries was saved in phys object)
...