Process control engineer Guy Dumont and pediatric anesthesiologist Mark Ansermino address the flood of information facing anesthesiologists in the operating room.
At first glance, pulp and paper and anesthesiology don’t seem to have much in common. But when process control engineer Guy Dumont (ECE), Director of UBC’s Pulp and Paper Centre, co-organized a Peter Wall Institute Exploratory Workshop on automation and robotics in health care in 2002, he knew what he was doing. There he met Mark Ansermino (Anesthesiology), a pediatric anesthesiologist and Director of Research for Pediatric Anesthesiology at BC Children’s Hospital. Since then, the two have made important strides in marrying Dumont’s signal processing expertise with Ansermino’s knowledge of physiological monitoring. The end goal is increased safety for patients under anesthesia.
Plumbing the Depths of Hypnosis
In one of their first projects, the researchers developed a technique for assessing the depth of hypnosis of patients under anesthesia. Wavelets with known properties are combined with the patient’s EEG signal to extract information from that signal. “We had been using wavelets for a long time in paper-machine control,” Dumont says, “and applied that experience to develop a depth-of-hypnosis sensor.” The system responds much more quickly than conventional systems, and is being commercialized by Cleveland Medical Devices. It also established the basis for a current collaboration on automated, closed-loop control of drug delivery in anesthesia, with control theorist Meeko Oishi (ICICS/ECE), and Bernard MacLeod and Stephan Schwarz from Anesthesiology.Good Vibrations
Anesthesiologists in a modern operating room are deluged with sensor data about their patient’s status. Threshold violations in exhaled gas levels, oxygen saturation, ventilation rate, heart rate, blood pressure, etc., are communicated through auditory or visual alarms. Many of these alarms are false, however, triggered by artifacts or signalling only a temporary, clinically insignificant deviation. “Auditory alarms are so frequent and false,” Ansermino says, “that you just don’t hear them anymore.” The researchers decided to try communicating alarms through tactile means, or the sense of touch, which had not been done before in medicine.With the help of human–computer interaction specialist Sid Fels (ICICS/ECE), they have devised a vibrotactile belt that communicates clinically significant information to the anesthesiologist through different patterns of vibration, which cannot be ignored. Thus, clinicians can keep their attention focused on the patient while receiving the information, instead of looking at a visual monitor. In an initial test of the belt involving a patient simulator, vibrating motors on the right side of the belt indicated significant changes in peak airway pressure, and on the left in the volume of air exhaled in one minute. Clinicians wearing the belt correctly diagnosed and responded to a simulated case of anaphylaxis more rapidly than those in the control group.
Encouraged, Dumont and Ansermino produced a wireless prototype incorporating change-detection algorithms that signal an alarm only after a certain change in the trend of a variable has been observed, rather than a threshold being exceeded. Variations among patients and over the course of the operation can be accounted for without false alarms being triggered. Other algorithms developed by Dumont and his students filter out artifacts from electrocautery devices, and safety features alert the clinician when the device loses communication with a sensor. In a recent real-time operating room test, anesthesiologists were able to correctly decode four different vibration patterns signalling the parameter being monitored, and the level and direction of change. “The belt is a less intrusive method of sending information to the clinician,” Ansermino says. “It’s a human–machine interface that gets to the subconscious level of communication.” He and Dumont will demonstrate the device at the American Society of Anesthesiologists’ annual meeting in October 2009.
Capturing Expert Knowledge for Decision Support
Expert systems that intelligently interpret data to provide real-time decision support are well established in the aviation and atomic energy industries. They are not used in medicine, however, and Dumont and Ansermino believe anesthesiology is a good place to start. They have developed a system that combines their changedetection algorithms with a rule engine based on practitioner consensus. For example, an anesthetized patient’s heart rate may go up because they are insufficiently anesthetized, or are bleeding. A rule can be added to the monitoring system based on the heart rate and blood pressure: a decrease in blood pressure with an increase in heart rate indicates the patient is likely bleeding, while an increase in blood pressure suggests the anesthesia is light. By building expert consensus on critical levels, rules such as this can be incorporated into the system to trigger appropriate real-time auditory, visual, or tactile alarms, as well as just-in- time information. The researchers are currently focusing on ventilatory events such as changes in lung compliance, as they are critical and can be missed by clinicians.Dumont and Ansermino envision an overall monitoring system that incorporates feature extraction, refined auditory, visual, and tactile communication, expert knowledge, and automated drug delivery for use in and beyond anesthesiology, such as in ICU bedside monitoring. Dumont’s process control expertise combined with Ansermino’s clinical experience makes theirs an ICICS collaboration made in heaven, or at least at the Peter Wall Institute.
Guy Dumont can be reached at 604-822-8564 or guyd@ece.ubc.ca, and Mark Ansermino at 604-875-2711 or mansermino@cw.bc.ca
