This Annex was formerly located in Annex J of PS 3.3 in the 2003 and earlier revisions of the standard.
Waveform acquisition is part of both the medical imaging environment and the general clinical environment. Because of its broad use, there has been significant previous and complementary work in waveform standardization of which the following are particularly important:
ASTM E31.16 - E1467 Specification for Transferring Digital Neurophysiogical Data Between Independent Computer Systems
CEN TC251 PT5-007 - prENV1064 draft Standard Communications Protocol for Computer-Assisted Electrocardiography (SCP-ECG).
CEN TC251 PT5-021 - draft Vital Signs Information Representation Standard (VITAL)
HL7 Automated Data SIG - HL7 Version 2.3, Chapter 7.14-20
IEEE P1073 - draft Medical Information Bus Standard (MIB)
DICOM - NEMA PS3.3, Section A.10 Standalone Curve Information Object Definition
For DICOM, the domain of waveform standardization is waveform acquisition within the imaging context. It is specifically meant to address waveform acquisitions which will be analyzed with other data which is transferred and managed using the DICOM protocol. It allows the addition of waveform data to that context with minimal incremental cost. Further, it leverages the DICOM persistent object capability for maintaining referential relationships to other data collected in a multi-modality environment, including references necessary for multi-modality synchronization.
Waveform interchange in other clinical contexts may use different protocols more appropriate to those domains. In particular, HL7 may be used for transfer of waveform observations to general clinical information systems, and MIB may be used for real-time physiological monitoring and therapy.
The waveform information object definition in DICOM has been specifically harmonized at the semantic level with the HL7 waveform message format. The use of a common object model allows straightforward transcoding and interoperation between systems that use DICOM for waveform interchange and those that use HL7, and may be viewed as an example of common semantics implemented in the differing syntaxes of two messaging systems.
Note: HL7 allows transport of DICOM SOP Instances (information objects) encapsulated within HL7 messages. Since the DICOM and HL7 waveform semantics are harmonized, DICOM Waveform SOP Instances need not be transported as encapsulated data, as they can be transcoded to native HL7 Waveform Observation format.
The following are specific use case examples for waveforms in the imaging environment.
Case 1: Catheterization Laboratory - During a cardiac catheterization, several independent pieces of data acquisition equipment may be brought together for the exam. An electrocardiographic subsystem records surface ECG waveforms; an X-ray angiographic subsystem records motion images; a hemodynamic subsystem records intracardiac pressures from a sensor on the catheter. These subsystems send their acquired data by network to a repository. These data are assembled at an analytic workstation by retrieving from the repository. For a left ventriculographic procedure, the ECG is used by the physician to determine the time of maximum and minimum ventricular fill, and when coordinated with the angiographic images, an accurate estimate of the ejection fraction can be calculated. For a valvuloplasty procedure, the hemodynamic waveforms are used to calculate the pre-intervention and post-intervention pressure gradients.
Case 2: Electrophysiology Laboratory - An electrophysiological exam will capture waveforms from multiple sensors on a catheter; the placement of the catheter in the heart is captured on an angiographic image. At an analytic workstation, the exact location of the sensors can thus be aligned with a model of the heart, and the relative timing of the arrival of the electrophysiological waves at different cardiac locations can be mapped.
Case 3: Stress Exam - A stress exam may involve the acquisition of both ECG waveforms and echocardiographic ultrasound images from portable equipment at different stages of the test. The waveforms and the echocardiograms are output on an interchange disk, which is then input and read at a review station. The physician analyzes both types of data to make a diagnosis of cardiac health.
Synchronization of acquisition across multiple modalities in a single study (e.g., angiography and electrocardiography) requires either a shared trigger, or a shared clock. A Synchronization Module within the Frame of Reference Information Entity specifies the synchronization mechanism. A common temporal environment used by multiple equipment is identified by a shared Synchronization Frame of Reference UID. How this UID is determined and distributed to the participating equipment is outside the scope of the standard.
The method used for time synchronization of equipment clocks is implementation or site specific, and therefore outside the scope of this proposal. If required, standard time distribution protocols are available (e.g., NTP, IRIG, GPS).
An informative description of time distribution methods can be found at: http://www.bancomm.com/cntpApp.htm
A second method of synchronizing acquisitions is to utilize a common reference channel (temporal fiducial), which is recorded in the data acquired from the several equipment units participating in a study, and/or which is used to trigger synchronized data acquisitions. For instance, the “X-ray on” pulse train which triggers the acquisition of frames for an X-ray angiographic SOP Instance can be recorded as a waveform channel in a simultaneously acquired hemodynamic waveform SOP Instance, and can be used to align the different object instances. Associated with this Supplement are proposed coded entry channel identifiers to specifically support this synchronization mechanism (DICOM Terminology Mapping Resource Context Group ID 3090).
Figure C.4-1 shows a canonical model of waveform data acquisition. A patient is the subject of the study. There may be several sensors placed at different locations on or in the patient, and waveforms are measurements of some physical quality (metric) by those sensors (e.g., electrical voltage, pressure, gas concentration, or sound). The sensor is typically connected to an amplifier and filter, and its output is sampled at constant time intervals and digitized. In most cases, several signal channels are acquired synchronously. The measured signal usually originates in the anatomy of the patient, but an important special case is a signal which originates in the equipment, either as a stimulus, such as a cardiac pacing signal, as a therapy, such as a radio frequency signal used for ablation, or as a synchronization signal.
Figure C.4-1 - Waveform Acquisition Model
The part of the composite information object which carries the waveform data is the Waveform Information Entity (IE). The Waveform IE includes the technical parameters of waveform acquisition and the waveform samples.
The information model, or internal organizational structure, of the Waveform IE is shown in Figure C.5-1. A waveform information object includes data from a continuous time period during which signals were acquired. The object may contain several multiplex groups, each defined by digitization with the same clock whose frequency is defined for the group. Within each multiplex group there will be one or more channels, each with a full technical definition. Finally, each channel has its set of digital waveform samples.
Figure C.5-1 DICOM Waveform Information Model
This Waveform IE definition is harmonized with the HL7 waveform semantic constructs, including the channel definition attributes and the use of multiplex groups for synchronously acquired channels. The use of a common object model allows straightforward transcoding and interoperation between systems that use DICOM for waveform interchange and those that use HL7, and may be viewed as an example of common semantics implemented in the differing syntaxes of two messaging systems.
This section describes the congruence between the DICOM Waveform IE and the HL7 version 2.3 waveform message format (see HL7 version 2.3 Chapter 7, sections 7.14 – 7.20).
Waveforms in HL7 messages are sent in a set of OBX (Observation) Segments. Four subtypes of OBX segments are defined:
- The CHN subtype defines one channel in a CD (Channel Definition) Data Type
- The TIM subtype defines the start time of the waveform data in a TS (Time String) Data Type
- The WAV subtype carries the waveform data in an NA (Numeric Array) or MA (Multiplexed Array) Data Type (ASCII encoded samples, character delimited)
- The ANO subtype carries an annotation in a CE (Coded Entry) Data Type with a reference to a specific time within the waveform to which the annotation applies
Other segments of the HL7 message definition specify patient and study identification, whose harmonization with DICOM constructs is not defined in this Annex.
The Waveform Module Channel Definition sequence attribute (003A,0200) is defined in harmonization with the HL7 Channel Definition (CD) Data Type, in accordance with the following Table. Each Item in the Channel Definition sequence attribute corresponds to an OBX Segment of subtype CHN.
Table C.6-1Correspondence Between DICOM and HL7 Channel Definition
|DICOM Attribute||HL7 CD Data Type Component|
|Waveform Channel Number||(003A,0202)||Channel Identifier (number&name)|
|Channel Source Sequence||(003A,0208)||Waveform Source|
|Channel Source Modifier Sequence||(003A,0209)|
|Channel Sensitivity||(003A,0210)||Channel Sensitivity and Units|
|Channel Sensitivity Units Sequence||(003A,0211)|
|Channel Sensitivity Correction Factor||(003A,0212)||Channel Calibration Parameters|
|Channel Time Skew||(003A,0214)|
|[Group] Sampling Frequency||(003A,001A)||Channel Sampling Frequency|
|Channel Minimum Value||(5400,0110)||Minimum and Maximum Data Values|
|Channel Maximum Value||(5400,0112)||(minimum&maximum)|
|Channel Offset||(003A,0218)||not defined in HL7|
|Filter Low Frequency||(003A,0220)|
|Filter High Frequency||(003A,0221)|
|Notch Filter Frequency||(003A,0222)|
|Notch Filter Bandwidth||(003A,0223)|
In the DICOM information object definition, the sampling frequency is defined for the multiplex group, while in HL7 it is defined for each channel, but is required to be identical for all multiplexed channels.
Note that in the HL7 syntax, Waveform Source is a string, rather than a coded entry as used in DICOM. This should be considered in any transcoding between the two formats.
In HL7, the exact start time for waveform data is sent in an OBX Segment of subtype TIM. The corresponding DICOM attributes, which must be combined to form the equivalent time string, are:
|Multiplex Group Time Offset||(0018,1068)|
The DICOM binary encoding of data samples in the Waveform Data attribute (5400,1010) corresponds to the ASCII representation of data samples in the HL7 OBX Segment of subtype WAV. The same channel-interleaved multiplexing used in the HL7 MA (Multiplexed Array) Data Type is used in the DICOM Waveform Data attribute.
Because of its binary representation, DICOM uses several data elements to specify the precise encoding, as listed in the following Table. There are no corresponding HL7 data elements, since HL7 uses explicit character-delimited ASCII encoding of data samples.
|Number of Waveform Channels||(003A,0005)|
|Number of Waveform Samples||(003A,0010)|
|Waveform Bits Stored||(003A,021A)|
|Waveform Bits Allocated||(5400,1004)|
|Waveform Sample Interpretation||(5400,1006)|
|Waveform Padding Value||(5400,100A)|
In HL7, Waveform Annotation is sent in an OBX Segment of subtype ANO, using the CE (Coded Entry) Data Type CE. This corresponds precisely to the DICOM Annotation using Coded Entry Sequences. However, HL7 annotation ROI is to a single point only (time reference), while DICOM allows reference to ranges of samples delimited by time or by explicit sample position.
The SCP-ECG standard is designed for recording routine resting electrocardiograms. Such ECGs are reviewed prior to cardiac imaging procedures, and a typical use case would be for SCP-ECG waveforms to be translated to DICOM for inclusion with the full cardiac imaging patient record.
SCP-ECG provides for either simultaneous or non-simultaneous recording of the channels, but does not provide a multiplexed data format (each channel is separately encoded). When translating to DICOM, each subset of simultaneously recorded channels may be encoded in a Waveform Sequence Item (multiplex group), and the delay to the recording of each multiplex group shall be encoded in the Multiplex Group Time Offset (0018,1068).
The electrode configuration of SCP-ECG Section 1 may be translated to the DICOM Acquisition Context (0040,0555) sequence items using DICOM Terminology Mapping Resource Template 3401 and Context Groups 3263 and 3264.
The lead identification of SCP-ECG Section 3, a term coded as an unsigned integer, may be translated to the DICOM Waveform Channel Source (003A,0208) coded sequence using Context Group 3001.
Pacemaker spike records of SCP-ECG Section 7 may be translated to items in the Waveform Annotations Sequence (0040,B020) with a code term from Context Group 3335. The annotation sequence item may record the spike amplitude in its Numeric Value and Measurement Units attributes.