RTG Home  |   PRECISE  |   CIS Home  |   Penn Engineering  |   Penn
Parameter-invariant Monitors for Medical CPS


Overview

With recent advances in low-power low-cost communication, sensing, and actuation technologies, Medical Cyber Physical Systems (MCPS) have revolutionized automated medical diagnostics and care. With this revolution, dawns a new era of medical monitoring where fusing measurements from multiple devices provides unprecedented early detection of critical conditions. However, often explicit models and/or rich training data relating available measurements to the critical conditions are unavailable or impractical. Under these troublesome scenarios, a parameter invariant approach to medical monitor design has been successful in developing monitors for conditions related to hypoxia, diabetes, and hypovolemia. Owing its mathematical origin to the robust radar signal processing literature, the parameter-invariant approach to medical monitor design is presented as consisting of three parts: (1) monitoring problem design, (2) low-order physiological modeling, and (3) constant false alarm rate (CFAR) testing.

The monitoring problem design overviews of the parameter-invariant approach and discusses how to design a robust medical monitoring problem statement. The first step in designing any monitor requires specifying objective in the form of a question about a past (or current) event. However, many alarm-worthy medical conditions are questions about future events. For instance, when monitoring hypoglycemia (critically low blood sugar) in type I diabetics, we demonstrate how the parameter-invariant approach aims to test whether the patient needs to eat, rather than testing for hypoglycemia directly. Monitors related to hypoxia, hypovolemia, and hypoglycemia have been developed in a series of case studies.

While there are many approaches to physiological modeling, the parameter-invariant approaches utilizes a compartment-modeling approach to develop low-order models which accurately describe physiological trends, subject to noise uncertainty. Examples illustrating how to identify these trends from published medical studies and patient physiological data are provided. The aforementioned case studies are employed to reinforce the usefulness of this approach, including the modeling of uncertainty of the general trend using a noise of unknown variance.

Through model manipulation and extensive use of null space projections, a CFAR test for the critical event is designed providing near-constant performance across the population. This is achievable by first generating a statistic constrained to a class of parameter-invariant statistics, then designing a test that simultaneously monitors the patient condition while ensuring the model accuracy and testing power is sufficient.

Case Studies

Non-invasive monitoring of hypovolemia

Hypovolemia caused by internal hemorrhage is a major cause of death in critical care patients. However, hypovolemia is difficult to diagnose in a timely fashion, as obvious symptoms do not manifest until patients are already nearing a critical state of shock. Novel non-invasive methods for detecting hypovolemia in the literature utilize the photoplethysmogram (PPG) waveform generated by the pulse-oximeter attached to a finger or ear. Until now, PPG-based alarms have been evaluated only on healthy patients under ideal testing scenarios (e.g., motionless patients); however, the PPG is sensitive to patient health and significant artifacts manifest when patients move. Since patient health varies within the intensive care unit (ICU) and ICU patients typically do not remain motionless, this work introduces a PPG-based monitor designed to be robust to waveform artifacts and health variability in the underlying patient population. To demonstrate the promise of our approach, we evaluate the proposed monitor on a small sample of intensive care patients from the Physionet database. The monitor detects hypovolemia within a twelve hour window of nurse documentation of hypovolemia when it is present, and achieves a low false alarm rate over patients without documented hypovolemia.

Monitors for meal detection in artificial pancreas

Blood glucose management systems are an important class of Medical Cyber-Physical Systems that provide vital everyday decision support service to diabetics. An artificial pancreas, which integrates a continuous glucose monitor, a wearable insulin pump, and control algorithms running on embedded computing devices, can significantly improve the quality of life for millions of Type 1 diabetics. A primary problem in the development of an artificial pancreas is the accurate detection and estimation of meal carbohydrates, which cause significant glucose system disturbances. Meal carbohydrate detection is challenging since post-meal glucose responses greatly depend on patient-specific physiology and meal composition.

We develop a novel meal-time detector that leverages a linearized physiological model to realize a (nearly) constant false alarm rate (CFAR) performance despite unknown model parameters and uncertain meal inputs. Insilico evaluations using 10, 000 virtual subjects on an FDA-accepted maximal physiological model illustrate that the proposed CFAR meal detector significantly outperforms a current state-of-the-art meal detector that utilizes a voting scheme based on rate-of-change (RoC) measures. The proposed detector achieves 99.6% correct detection rate while averaging one false alarm every 24 days (a 1.4% false alarm rate), which represents an 84% reduction in false alarms and a 95% reduction in missed alarms when compared to the RoC approach.

Early Detection of Critical Pulmonary Shunts in Infants

This project aims to improve the design of modern Medical Cyber Physical Systems through the addition of supplemental noninvasive monitors. Specifically, we focus on monitoring the arterial blood oxygen content (CaO2), one of the most closely observed vital signs in operating rooms, currently measured by a proxy - peripheral hemoglobin oxygen saturation (SpO2). While SpO2 is a good estimate of O2 content in the finger where it is measured, it is a delayed measure of its content in the arteries. In addition, it does not incorporate system dynamics and is a poor predictor of future CaO2 values. Therefore, as a first step towards supplementing the usage of SpO2, this work introduces a predictive monitor designed to provide early detection of critical drops in CaO2 caused by a pulmonary shunt in infants.

To this end, we develop a formal model of the circulation of oxygen and carbon dioxide in the body, characterized by unknown patient-unique parameters. Employing the model, we design a matched subspace detector to provide a near constant false alarm rate invariant to these parameters and modeling uncertainties. Finally, we validate our approach on real-patient data from lung lobectomy surgeries performed at the Children's Hospital of Philadelphia. Given 198 infants, the detector predicted 81% of the critical drops in CaO2 at an average of about 65 seconds earlier than the SpO2-based monitor, while achieving a 0:9% false alarm rate (representing about 2 false alarms per hour).

Publications

TUTORIAL ON PARAMETER-INVARIANT MONITORING

Tutorial slides presented at ESWEEK 2015:

Tutorial slides presented at ICHI 2015, with more emphasis of medical applications:


RTG Home  |   PRECISE  |   CIS Home  |   Penn Engineering  |   Penn