Physiologic Datasets (Waveform and Numerics)

Datasets can be single channels or they can be a table of multiple channels.

Single Column Datasets

In this case, each signal is stored in a separate dataset in its respective group. The file structure is:

/ (root)
.meta

mapping

/ numerics
.meta
/HR
/ABP-S
/ABP-D
/ABP-M
/SPO2

/ waveforms
.meta
/ECG-II
/ABP
/CVP

/ clinial
/notes
/labs
.test_info
/micro
.test_info
/infusions
/diagnosis
.coding

Tabular Datasets

Tabular datasets will generally be used when multiple similar signals (generally same sample rate and source) are stored together. The resulting file structure looks like:

/ (root)
.meta

mapping

/ numerics
/vitals
.meta

/ waveforms
/hemodynamics
.meta

/ clinial
/notes
/labs
.test_info
/micro
.test_info
/infusions
/diagnosis
.coding

Physiologic Dataset Metadata

Dataset Metadata Fields

physiologic dataset metadata (eg: /waveforms/ABP/.meta)

These datasets can be single or multicolumn but the metadata attribute will be similar in both cases. Timestamps may be stored as a number of seconds from the origin (base_datetime) or they may be implied. In the case of implied timestamps, the timeseries is reconstructed using the sample_rate and the row number. Single signal datasets can be further divided into two types:

  • with timestamps (compound dtype)

  • with implied timestamps (float dtype)

All datasets should include at at least a .meta attribute containing time orign and sample rate fields. If the datasets do not have this information in the .meta attribue then the time_origin will be assumed to be that stored in the metadata for the containing group (eg waveforms or numerics) or in the root group.

sample_rate

sample rate in Hz

time_origin

the time origin of the dataset formatted as “YYYY-MM-DD HH:MM:SS.ssss”

Additional information (particularly for tabular datasets) is included in the column field:

dataset column metadata (eg: /waveforms/ABP/.meta)
Parameters
  • type (str) – column type {time or real}

  • LOINC (str) – LOINC identifier for a signal column

  • uom (str) – units of measurement

  • scale (float) – scale factor [optional] - used to convert raw data stored as int in the dataset to physical units (this will be set to 1 or omitted entirely if the data is already stored as float)

  • fmt – format [optional]

Sample dataset metadata stored as JSON formatted attribute as shown below.

/group/dataset/.meta
    {
        "time_origin": "2116-12-24 12:35:06.147000",
        "sample_rate": 0.0166666666667,
         "columns": {
             "HR": {
             "type": "real",
             "LOINC": "8867-4",
             "uom": "bpm",
             "scale": 1,
             "fmt": "16",
             "baseline": 0
            },
             "CVP": {
                 "type": "real",
                 "LOINC": "60985-9",
                 "uom": "mmHg",
                 "scale": 1,
                 "fmt": "16",
                 "baseline": 0
            },
             "SpO2": {
             "type": "real",
             "LOINC": "",
             "uom": "%",
             "scale": 1,
             "fmt": "16",
             "baseline": 0
            },
        "NBPSys": {
            "type": "real",
            "LOINC": "76534-7",
            "uom": "mmHg",
            "scale": 1,
            "fmt": "16",
            "baseline": 0
            },
        }