Introduction

Long-term remote medical monitoring has the potential to improve the quality of healthcare for all individuals. Patients suffering from chronic conditions are likely to benefit the most. By continuously monitoring chronically ill patients, physicians will be able to better understand, treat, and manage these patients. Moreover, continuous monitoring offers physicians the opportunity to collect physiological data about a patient throughout his daily routine. As an additional benefit, pervasive medical monitoring can discover latent diseases sooner without hospitalizing a patient or requiring frequent visits to a physician's office.

System Architecture

To effectively monitor patients, a pervasive medical monitoring system must be able to follow a subject throughout her daily routine. In our medical monitoring system, we facilitate this mobility by attaching wireless medical sensors to the subject. Our initial design supports a single subject and consists of five main components: sensors, sensor proxies, a real-time streaming database with long-term storage capabilities, and communication facilities.

Unlike traditional medical sensors, each body-worn sensor consists of a local wireless radio, simple computation elements, and sampling hardware in addition to the sensor itself. The wireless medical sensors in our system sample physiological data from the subject and wirelessly transmit the data to a sensor proxy.

Sensor proxies can be implemented on portable devices with a wireless network connection such as a cell phone or WiFi-enabled PDA. The sensor proxies translate the data into streams of timestamped data and send the streams over the network (or Internet) to a streaming database [1]. The database processes the data in order to recognize user behavior, monitor physiological state, and detect anomalous medical conditions. The output of the detection algorithms can be used to raise alerts and summon caregivers when an emergency occurs.

The Energy Consumption Problem

In order to build a successful pervasive sensor network for chronic illness management, the energy consumed by the untethered smart sensors and portable devices in the system must be minimized as much as possible without compromising the ability of the detection algorithms to detect and diagnose disease.

In our proposed system, our sensors are augmented with computation, communication, and storage capabilities. This augmentation allows detection algorithms to distribute their computation. In addition, the algorithms within the system can adapt the sensors to the patient and particular illness as needed.

Because our sensors are small, they are extremely energy constrained. Therefore, we must pay particular attention to their energy consumption. A typical sensor in our system will consist of four main components: a sampling component, a computation component, a storage component, and a communication component. The energy consumed by each sensor can be modelled using the following equation:

Since each component has a fixed power consumption depending on its mode operation, we can further decompose the equation above:

Here, we have assumed that each component can operate in either active mode or idle mode and that there are n sensors in our system.

In a naïve medical monitoring system, the smart sensor might be configured to simply sample and transmit data continuously. Configured in this fashion, our smart sensors will drain their batteries quickly. A well-known technique to reduce the energy consumed is to perform computation at the sensor in order to reduce the amount of data transmitted. For example, we could transmit less data by compressing data at the sensor. Since the energy of transmission is much greater than the energy of computation, by trading a little computation for communication, the lifetime of the sensor can be increased dramatically.

Performing compression at each sensor is only one way to reduce the energy consumption of the pervasive medical monitoring system. Our research will explore some novel energy saving techniques that can be used to extend the overall lifetime of a remote medical monitoring system. In particular, we are investigating ways to intelligently duty cycle the components in our system such that important events are not missed while the overall lifetime of the sensor is increased. Some of the predictive shutdown techniques developed to save energy for laptop computers [2] may provide us with systematic duty cycling techniques to use.

Current Progress

Currently, we are in the process of modeling the components (hardware and software) of our system. By providing a mode of a monitoring system, we hope to develop a framework to compare various energy saving algorithms. We hope that this framework can then be used by medical detection algorithm designers who may wish to deploy their algorithms in an energy constrained environment. In addition to building a framework, we hope to implement a few of the algorithms in a real medical monitoring context.

Research Support

This research is supported by the National Library of Medicine, Acer Inc., Delta Electronics Inc., HP Corp., NTT Inc., Nokia Research Center, and Philips Research under the MIT Project Oxygen Partnership, and CIMIT, the Center for the Integration of Medicine and Innovative Technology.

References

[1] Daniel J. Abadi, Don Carney, Ugur Çetintemel, Mitch Cherniack, Christian Convey, Sangdon Lee, Michael Stonebraker, Nesime Tatbul, and Stan Zdonik. Aurora: a new model and architecture for data stream management. The VLDB Journal, pp. 120-139, Vol. 12, No. 2, 2003.

[2] Mani B. Srivastava, Anantha P. Chandrakasan, and Robert W. Brodersen. Predictive System Shutdown and Other Architectural Techniques for Energy Efficient Programmable Computation. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, pp. 42-55, Vol. 4, No. 1, March 1996.