ICU Patient Deterioration Prediction : A Data-Mining Approach
Published Research Paper
1.
I. BACKGROUND
Automation of pump control by closed-loop systems has
long-recognized advantages. The attention of the clinician may
be directed to other matters, while desired conditions are
automatically maintained. Such systems been explored for
over 20 years [1, 2]. In some cases they have demonstrated
performance superior to expert clinicians [3]. Potential hazards
are also recognized and must be addressed in model systems.
A critical enabling technology for such closed-loop
systems is open nonproprietary standard interfaces for
acquisition and transmission of data in a device network
facilitating monitoring and control, allowing monitors,
sensors, pumps, and controllers to be integrated. Efforts at
standard device interfaces over the past 30 years [4, 5], have,
however, borne little fruit. Only a few proprietary or
customized systems have been developed. Enabling
nonproprietary device interactions has been a pipedream. Yet,
this is essential to enable research, development, and adoption
of systems of interconnected devices.
II. PURPOSE
The main goal of this work is to develop minimalist open
standard interfaces to facilitate research and development.
Clinically-appropriate standardization would require much
more sophistication. Standardization enables interaction of
components with otherwise incompatible proprietary
communications. Open software in the device interfaces and
protocols permits logging of communication at every
interface level to fully understand all interactions of closed-
loop systems. Applications include capture of clinical events
and interventions, and implementation of clinical controller
systems. Work in progress includes development of closed-
loop physiological animal testing and computer-in-the-
middle model systems.
III. METHODS
We have adopted hardware and an architecture similar to the
integrated clinical environment work by MDPnP [5]. Each
medical device is connected to a dongle that translates the
specific proprietary device protocol into a standardized one.
The dongles then connect to a local network allowing them to
interact with applications on a computer that handle device
and data management and processing algorithms. Unlike the
* Research in part by Bucknell-Geisinger Research Initiative Phase IV
Grant No. GPE094
P. Asare, A. Acharya, Y. Huang, D. Karki, W. Kyaw, C. Mahoney are
with Bucknell University, Lewisburg, PA 17837 USA
MDPnP work, Python is used for software development to
allow rapid-prototyping and ease of adoption and
modification. In addition, the Robot Operating System (ROS)
[6] is our middleware because of its proven effectiveness for
developing production-grade distributed systems and Python
integration.
IV. RESULTS
We will demonstrate a system that: (a) interacts with a patient
monitor (Philips IntelliVue MP50) and sends commands to an
unmodified commercial pump (CME America BodyGuard
121 Dual-Channel Infusion Pump); (b) detects whether a
device is connected to the system; (c) logs system
performance, device data, and monitored data (simulated
patient data and infusion actions); (d) configures and controls
the system through a graphical user interface; (e) runs a
simple illustrative closed-loop control scenario.
V. CONCLUSION
Our progress suggests that an easier-to-use experimentation
platform for closed-loop control is feasible.
ACKNOWLEDGMENT
The authors thank CME America, LLC for providing us
with the infusion pump.
REFERENCES
[1] D. Ramakrishna, K. Behbehani, K. Klein, J. Mokhtar, W.W. von
Maltzahn, R.C. Eberhart, M.Dollar, “In vivo evaluation of a closed loop
monitoring strategy for induced paralysis.” J Clin Monit Comput. 1998
Aug;14 (6):393–402.
[2] G. C. Kramer, M. P. Kinsky, D. S. Prough, J. Salinas, J. L. Sondeen, M.
L. Hazel-Scerbo, C. E. Mitchell. “Closed-loop control of fluid therapy
for treatment of hypovolemia.” J Trauma. 2008 Apr;64(4 Suppl):S333-
41.
[3] J. Rinehart, M. Lilot, C. Lee, A. Joosten, T. Huynh, C. Canales, D.
Imagawa, A. Demirjian, M. Cannesson, “Closed-loop assisted versus
manual goal-directed fluid therapy during high-risk abdominal surgery:
a case–control study with propensity matching.” Crit Care [Internet].
2015 19(1).
[4] R. J. Kennelly, R. M. Gardner, “Perspectives on development of IEEE
1073: the Medical Information Bus (MIB) standard.” Int J Clin Monit
Comput. 1997 Aug;14(3):143-9.
[5] Medical Device "Plug-and-Play" Interoperability Program.
http://mdpnp.org
[6] Robot Operating System (ROS). http://www.ros.org/.
(P. Asare is corresponding author: phone: 570-577-2344; fax: 570-577-1449;
e-mail: philip.asare@bucknell.edu).
S. M. Poler, J. R. La Valley, R. Tevis are with Geisinger Medical Center,
Geisinger Health System, Danville, PA 17822 USA.
Demo of Platform for Enabling Research and Development of
Closed-Loop Control of Infusion in the Operating Room*
Philip Asare, Member, IEEE, Adit Acharya, Yuxuan Huang, Dikendra Karki, Win Kyaw, Caitlin
Mahoney, S. Mark Poler, Senior Member, IEEE Jean R. La Valley, Rick Tevis