PORTABLE PHYSIOLOGICAL SIGNAL
PROCESSOR
FOR
A NOVEL PROGRAMMABLE
CARDIAC MONITOR
PURPOSE OF THE PROJECT
The Biomedical
Engineering Department of the Johns Hopkins University and DVP,
Inc. of Rockville, MD, now InHand
Electronics, Inc., have successfully competed and received a Phase
II SBIR award from the NIH. The Phase II award to DVP was for developing
a Portable Physiological Signal Processor (PSP), a novel device
that allows flexible, programmable interface for digital signal processing
in a portable unit.
The main purpose of the present project, supported by Maryland Industrial
Partnerships Program (MIPS, Contract Agreement # 2209) is to create
an innovative and commercially attractive product for cardiac monitoring.
It is anticipated that this leveraged funding will allow us undertake the
research on the new invention leading to a new product with exciting marketing
potential and pave way for a more rapid access to commercialization.
The specific goal of this project is to develop a versatile, easily
PROGRAMMABLE
CARDIAC MONITOR (PCM) to be used in cardiological practices
and for research. The project represents a collaboration between JHU and
DVP, Inc.( InHand Electronics,
Inc.). On behalf of Johns Hopkins, Dr.
Boris Gramatikov is the person responsible for hardware
and software development. Mr. Boris Donskoy
is in charge on the DVP side. Here are the main issues of the collaboration
project:
-
JHU will develop
ECG amplifiers and data acquisition, and will further research and develop
the concept of myocardial ischemia detection using wavelet analysis of
ECG signals. JHU will 1) develop a continuous wavelet transform analysis
algorithms for ECG signal processing, and 2) demonstrate two applications
of the PSP/PCM: i) detection of acute myocardial ischemia in patients presenting
to the Emergency Department with chest pain, and ii) detection of ischemia
during exercise stress testing.
-
DVP, Inc, will further develop the
PSP suitable for monitoring ECG signals and collaborate with JHU to implement
the ECG signal processing algorithms and implement the wavelet analysis
signal processing software. DVP will create a PSP prototype and share it
with JHU. DVP will also collaborate on actual prototype development for
the applications outlined above.
The two organizations will collaborate to commercialize the Programmable
Cardiac Monitor technology. The long term goal is to develop the PCM based
on the PSP technology [1,2]. This will lead to a generic signal processing
technology and an ECG processing technology. The PSP will provide the hardware
platform and signal processing tools. The PCM will provide ECG signal processing
software validated for different medical applications. Each application
does not necessarily depend on the others and may be financed and eventually
manufactured separately. The PSP is expected to increase the marketing
potential of the PCM.
A PSP prototype has already been developed [2]. The prototype included
a motherboard (incl. ADC, DAC), and analog daughterboard, graphic LCD display
(340x240 pixels) and a keypad. Two PCMCIA Type II slots are provided. One
PCMCIA slot is intended to accept memory application cards. Two full duplex
serial communication channels are provided. At present, work is aimed at
expanding and improving the existing resources, to create a powerful, turnkey
environment for the design of many different physiological instruments.
The generic core of the earlier version of the PSP was a 16-bit Hitachi
SH7032 RISC processor used as both a system controller and a signal processor
[1].
The company is developing an extended version by adding the next generation
of RISC technology [3], substantially increasing signal processing capabilities
over the existing device. Further, a significant data logging capability
is being added which will allow interface to a high capacity memory card
or an ATA-type disk drive (through the PCMCIA slot). Such peripherals are
currently commercially available in over 20 MB flash memories and 300 MB
hard disk drives. These devices are relatively low in cost and, due to
their current popularity, are expected to continue becoming more capable
and less expensive.
Hardware development
-
We studied the various possible options of the "daughter
board" to accomodate the ECG amplifiers, the data acquisition and transmission,
DC-DC converter and isolation .
-
We finalized the data acquisition and data transfer concept, based on the
22 bit Sigma-Delta analog-to-digital converter AD 7716.
-
We purchased an evaluation board AD7716EB, which helped us test the AD
converters and better design the data acquisition part of the daughter
board.
-
We completed the amplifier design, based on low-noise instrumentation amplifiers
AD 620 and the concept of "driven right leg".
-
The JHU partners learned how to work with ORCAD, designed the amplifiers
on the daughterboard, and suggested the analog-to-digital circuitry and
communication. The DVP team then introduced a series of improvements and
finalized the design of the daughterboard.
The Daughter Board design was done by
Dr.
B. Gramatikov, JHU, and DVP, Inc
(now InHand Electronics, Inc.),
working in close collaboration. The design underwent several iterations
on OrCad *.dsn file level. The board layout (6 layers, surface mounted
technology) was carried out by
Diversified Computer Services, Inc.,
who have specialized in multi-layer boards. After that, Quick Turn Circuits,
Inc did board manufacturing and testing. Then, each partner, JHU and
DVP, Inc, populated one daughter board and tested it.
Parallelizing of tasks: To avoid unnecessary delay
and to make the design more flexible, we decided to split the project into
two parallel paths:
a) at JHU:
-
development of a support front-end module containing the daughter card
(ECG amplifiers, data acquisition using the same AD converter 7716 as the
target application), and a communication card to exchange data and control
signals with a PC using its parallel port. Thus, software development using
the C-language (consistent and platform-independent) could be started well
ahead on a PC, before the new OmniMeter
[3] design was completed.
-
the front-end module also was used to test the DC-DC converter
-
The front-end module was placed in a plastic enclosure.
A standard Physio-Control ECG clinical grade patient cable was connected
to its input, whereas its parallel port was connected to a laptop PC. This
configuration was demonstrated during the technical review on April 14,
1999, in front of MIPS representatives. During the presentation, human
ECG data were acquired and displayed in real time. Also, the software (please
see next sub-section) was run on files from the ST-T Data Base.
b) at DVP, Inc:
-
design of a new, much more powerful hardware platform of the basic device,
the OmniMeter,
now based on a 200 MHz StrongARM
SA-1100 processor; this was a revolutionary improvement [3].
-
finalized the design of the daughter card, which now accommodates ECG amplifiers,
data acquisition, isolation, and fast communication with the main processor
of the Physiological Signal Processor. In the second half of April 1999,
the company tested some of the main functions of the daughter board in
the new device, including ECG amplification and data acquisition (analog-to-digital
conversion plus real time graphics display).
-
the new design of the Portable Physiological Signal Processor with the
newly developed daughter board was shown at the Sensors ExpoTM,
Baltimore, May 4 - May 6, 1999.
Software and algorithms design
a) The JHU partners at first developed the real-time
model of the Portable Ischemia Monitor as the first targeted application
of the Physiological Signal Processor. The software algorithms include
(but are not limited to) QRS identification, waves recognition, identification
and exclusion of ectopic beats, beat averaging, ST measurement, calculating
the Time-Frequency Distribution (TFD) of the signal during
the QRS complex by means of the Continuous Wavelet Transform,
calculating an appropriate QRS-based ischemia detection index, and a Graphical
User Interface (GUI).
b) The JHU partner wrote the Wavelet Transform algorithms
in the C-language and tested them on an SGI Indigo workstation. Originally
these algorithms had been implemented in Matlab at JHU. Porting onto the
Physiological Signal Processor requires C-programming to guarantee fast
operation.
c) The JHU partner created an algorithm for contour-plotting in C-language
and tested it on a workstation. This algorithm will eventually allow us
to plot the time-frequency distributions of the QRS complex on the screen
of the portable Physiological Signal Processor (during Phase II).
d) We purchased and installed a Microsoft Visual C++ 6.0 compiler
to allow emulation of the algorithms and programs on a PC. Later we will
port the software onto the Physiological Signal Processor. We do not anticipate
major problems at moving the software onto the OmniMeter, because of the
consistency and portability of C-language. The main signal processing algorithms
from (a) have been written in C-language and tested. Approximately 90%
of the software for morphological analysis of the time signal is finished.
e) With the aim of validating our concepts, we purchased the European
ST-T data base (as planned) and started analyzing the data in it.
This data base contains 2-hour ECG recordings of ambulatory patients with
transient episodes of "silent" (non-symptomatic) ischemia. It comes with
a convenient software library which helps the user read the accompanying
anotation files and localize precisely the ischemic episodes. So far we
have analyzed on a workstation ca. 50% of the data (41 patients), using
the wavelet transform algorithms written in C. The analysis confirmed
our assumptions that coronary ischemia causes intra-QRS high-frequency
changes due to altered cardiac depolarization front. These changes
accompany ST alterations and sometimes precede them.
f) To provide a user-friendly PC-based model
of the target device we developed a Graphical User Interface (GUI) version
of the data acquisition and analysis software using LabWindowsTM
(of National Instruments) available at the PI's lab. LabWindowsTM is a
versatile C++ language that incorporates many modern GUI features. The
LabWindowsTM based software prototype is
extremely helpful in many ways. So far it:
· enabled writing and testing of the low-level data acquisition
routines (AD conversion using the novel Sigma-Delta 22 bit 2 stage A-D
converter);
· helped develop different scroll charts and display graphs for
ECG signals and trends with results to be used in the target device;
· helped suggest appropriate functions that will be implemented
as touch-screen functions in the portable device;
· test analysis algorithms.
The LabWindowsTM
based software prototype was run together
with the front-end module. Its main role till now, however, was to test
the GUI and analysis algorithms. The software prototype shows the main
controls, panels and functions, to be implemented on the portable Ischemia
Monitor. Fifteen minutes of file e0103 of the ST-T data base were analyzed,
with a documented mild episode of ST changes and transient ischemia at
12 min after start (100 mV = 1 mm are considered the treshold value). The
ST trend plot clearly shows that, despite of a first peak which is an artefact.
The newly devised mid-QRS High-frequency Index trend plot responded significantly
better to this ischemic episode. The Time-Frequency distribution of the
averaged QRS will be optional in the target device.
Recently DVP, Inc. decided to use Windows
CE as a standard Operating System for the OmniMeter.
This not only improves the graphical user interface, but also creates endless
new opportunities such as adding TCP/IP interface (wireless or cable),
Internet, Data Base Capabilities, Microsoft Office etc.
Check Boris Gramatikov's Home Page, JHU
Go to DVP, Inc, or now
InHand Electronics, Inc.
Some related publications:
[1] Girson, A., Donskoy, B. Field-programmable instrument
highlights application potential of RISC embedded controllers.
EDN, July 3, 1997, pp. 105-107.
[2] Gendreau, P.R. Team creates PDA that morphs into
many variations. Portable
Desig, July 1998, pp.49-56.
[3] Gendreau, P.R. Spinning 2G configurable platforms
challenges designers. Portable
Desig, August 1999, pp.41-46.