Fig 1. An electrocardiogram comprises time-domain sweeps of the voltage generated by the heart muscle as it pumps. Four of the “leads” (traces) in this test image from a development system represent the differential signals from pairs of electrodes attached to different parts of the body. (Hence, there can be more leads than there are wired electrodes.) Looking at different leads gives cardiologists more information than they could get from a single lead. The green line here is a single-ended input from the “pace” electrode, which picks up the signal from the pacemaker if there is one. (courtesy of Analog Devices)
Fig 2. The functionality of the AFE for electrocardiographs is straightforward, except for the Wilson network. This resistor network sums the inputs from the patient’s two arms and left leg, providing a zero-voltage reference point at the center of the network. (courtesy of Maxim Integrated Products)
Fig 3. Pulse oximeters are based on capturing reflections of alternating red and infrared light flashes. The concentration of oxygen in the patient’s hemoglobin affects which wavelengths are more strongly reflected. (courtesy of Maxim)
Fig 4. The semiconductor content for most medical products intended for developing countries is frequently designed in North America and incorporated into finished products in China. Because of the high degree of integration, chip suppliers now have to provide complete reference designs, without which the optimum interconnection of the silicon products and the outside world would be far from obvious. This is an example from a Texas Instruments datasheet for a chipset reference design for an ultrasound device.
The “real world,” we say, is analog. And nowhere is it more important to keep the analog part real than in medical diagnostic equipment. Many common factors are related to digitizing analog inputs for electrocardiographs and other blood-related diagnostic gear, not to mention ultrasound imagers, blood-glucose meters and wearable insulin pumps, and digital thermometers. But there are unique constraints as well.
Blood and Circulation
Diagnostic instruments for blood-related problems include electrocardiographs (ECGs), pulse oximeters, and electronic blood-pressure measurement. Most of us are familiar with the function of ECGs, and many of us have had their wired electrodes taped about our bodies. But how they work may still be a mystery.
Heart signals, which have amplitudes of a few millivolts, can be picked up at many points on the body. By looking at the differential signals across different locations, physicians can assess heart functionality from multiple viewing angles. Reading those results is something of an art, because the signals are simply displayed as separate traces on the ECG display or printout. Each of those traces represents the differential voltage between two specific electrodes, or the differential voltage between one electrode and the average voltage from several others (Fig. 1).
By medical tradition, the channels are commonly referred to as “leads,” which is inconsistent with our engineering traditions and accounts for some confusion when the number of medical leads does not correspond to the number of input wires attached to the patient.
The maximum number of leads is usually 12, which means that, in general, a12-lead ECG has 10 electrodes. Nine of these monitor heart signals across different parts of the body. The tenth, called the “right leg” (RL) electrode, is electrically driven to reduce the common-mode voltage.
The nine input electrodes are the left arm (LA), right arm (RA), left leg (LL), and six precordial (chest) electrodes designated V1 through V6. When electrodes are grouped, their voltage is averaged. RA, LA, and LL are averaged for six of the leads and become one side of differential pairs with V1 to V6.
In addition, three of the leads measure the differential voltage between RA, LA, and LL and the average of the other two electrodes. The remaining three “leads” on a “12-lead” ECG readout represent combinations of RA, LA, and LL, measured differentially.
In addition to the body-generated signals, a higher-end ECG will have other input signals, such as the “pace” signal, which is generated by a pacemaker, if the patient has one. It is characterized by short duration (in the range of tens of microseconds to a few milliseconds, and an amplitude ranging from a few millivolts to 1 V). One of the challenges for engineers designing a new ECG is keeping the pace signal from interacting with the other channels.
Another common non-heart signal is called “lead-off,” and that is exactly the fault situation that it detects. To detect “lead-off,” the ECG measures the impedance between each of the differential-sensing electrodes and the lead-off electrode. Sometimes, this impedance measurement provides an input for respiration rate, because thoracic impedance changes as the chest rises and falls.
Now that we have straightened out leads and electrodes, the important analog element in the signal chain is the analog front end (AFE), including the analog-to-digital converter (ADC), that allows the rest of the system to make sense of those noisy differential signals (Fig. 2). The heartbeat signals of interest are in the millivolt range and cover a spectrum from 0.05 to 100 Hz, But the AFE also has to deal with pace and respiration signals.
Noise arises from cell phones, pace signals, lead-off signals, ac hum, and signals from other muscles besides the heart. Besides that, the various differential inputs can be riding on a dc offset of hundreds of millivolts, with channel-to-channel common-mode voltages of a volt or more. Defibrillation is another noise source that requires input circuit protection and fast recovery time.
Blood Pressure Monitors
Sphygmomanometers (blood pressure monitors) have fewer analog inputs than ECGs. In its most basic configuration, a sphygmomanometer comprises a calibrated air-inflatable cuff and a listening device. The cuff can go on either the upper arm or the wrist. For diagnostics, medical personel can use either a stethoscope to listen to arterial wall sounds (auscultatory method) or a pressure sensor to sense arterial wall vibrations (oscillometric method).
As most people have experienced, the cuff produces enough pressure to prevent blood flow in the arteries, and the pressure is gradually released until the blood starts to flow again. The pressure at which the flow starts to happen is the systolic pressure. The pressure at which there is no longer any restriction of blood flow is the diastolic pressure.
Auscultation requires a certain amount of skill and practice, which the oscillometric method replaces with the pressure sensor and electronics. In practice, the signal from the pressure sensor is conditioned with some sort of AFE and applied to an ADC. The systolic pressure, diastolic pressure, and pulse rate are then calculated in the digital domain.
Heart-rate fitness monitors are a subcategory of sphygmomanometer for non-clinical use, but that doesn’t make them simpler. Beyond sensing the heart rate (the cuff method of measuring blood pressure is not practical for instrumenting anyone performing a strenuous activity), they may also sense or measure temperature, respiration, and, for runners or cyclists, cadence.
For the various types of transducers involved, this is another application for AFEs and ADCs. But fitness monitors may also include timers, accelerometers, and even complete GPS systems. To avoid encumbering an athlete with wires, they may use a body-area network (BAN) such as Bluetooth to link sensors to the main monitoring unit.
Pulse oximeters optically measure the oxygen saturation of arterial blood, based on the light absorption characteristics of hemoglobin. At higher levels of oxygenation, hemoglobin absorbs more infrared light than red light, and vice versa. In the oximeter probe, which is affixed to a finger tip, red and infrared LEDs flash alternately, and a photodiode collects the light that is not absorbed (Fig. 3). In practice, it’s a little more complicated, as the instrument also has to account for pulsations in blood pressure.
As with many diagnostic medical instruments, there is a range of price/performance points. Higher-end, ac-powered, portable bedside monitors incorporate the optics at the end of a cable. They may display pulse rate and blood pressure as well as the level of oxygen. Mid-range models are typically battery-powered handheld units, and like the high-end units, the optics are remotely located at the end of a cable. Very low-end units are self contained.
At different points on the size scale, analog design differences are often dictated by battery size and power budgets. The red and infrared LEDs are pulsed slowly, usually at a rate less than 10 kHz, with an off-time in the duty cycle to allow ambient-light measurement.
Instrument designers cannot use LED drivers that are designed for backlighting or general illumination applications because they need more precise current control and because the switching regulators most commonly used create too much noise. In terms of the need for precision compared to ordinary LED drivers, designs typically use 10- or 12-bit digital-to-analog converters (DACs) driving transimpedance amplifiers.
On the receive side, a photodiode receives both ambient and modulated light from the two LEDs, and its output is converted to a voltage with a transimpedance amplifier. The challenge is to extract the information contained in the small amounts of red and infrared light in the presence of all that ambient light.
Fortunately, the ambient element can be removed by a high-pass filter, and the time scale is long, which favors the use of relatively slow but high-resolution ADCs. The key performance-limiting element is the transimpedance amplifier.
The fastest-growing diagnostics market in developing countries is in portable ultrasound (see “New Technology Treats Medical Needs In Developing Countries,” p. xx). Phased-array ultrasound systems can generate images of internal organs and structures, map blood flow and tissue motion, and provide blood-velocity information.
Ultrasound systems vary considerably in size and cost. The trend is toward higher integration in the probe and in the AFE and more sophisticated digital processing and displays.
The heaviest concentration of new designs for portable and handheld ultrasound units is in China, where North American and European companies competing for design wins have been positioning themselves on analog integration in terms of functionality (integrated AFE and ADC) and number of channels. In many cases, the Chinese companies concentrate on the manufacturing engineering and leave the functional design to outsiders (Fig. 4).
Transducers for ultrasound imaging systems are optimized for specific diagnostic applications. Essentially, the more transducers, the higher the resolution, but advances in software also enable designers to do more with less. Pulses are generated by anywhere from 32 to 256 piezoelectric elements, which operate at frequencies from 1 to 15 MHz. When the design involves fewer channels than transducers, it will include high-voltage switches to multiplex the signals.
The switches are a challenging design element. They have to handle transmit pulses with voltage swings as large as 200 V p-p with peak currents up to 2 A. They must switch rapidly to quickly modify the configuration of the active aperture and maximize image frame rate. Also, they must have minimal charge injection to avoid image artifacts.
In high-end systems, a transmit beamformer generates signals that are timed and phased to focus the signal from the transducer array. In operation, the beamformer generates words of 8 to 10 bits at rates up to 40 MHz to feed the DACs that produce the analog signals. The DAC outputs are amplified further to drive the transducers.
Lower-cost systems use multilevel high-voltage pulsers to generate the transmit signals, switching each transducer element to programmable high-voltage supplies to generate the transmit waveform. In the simplest of systems, a transmit pulser switches the element between positive and negative transmit supply voltages that are controlled by the digital beamformer. At the next higher level of complexity, the transducer elements are connected to multiple supplies with different voltage output levels (and to ground) to generate multilevel waveforms.
One factor complicating AFE design is second-harmonic imaging, a technique that takes advantage of the way the human body reflects much of the ultrasound energy that is applied. Essentially, energy is generated at a fundamental frequency at that frequency’s second harmonic. Ultrasound devices that process those higher-frequency signals show sharper images than devices that capture reflections of the fundamental. There are at least two ways ultrasound machines accomplish this.
In “standard-harmonic imaging,” the second harmonic is suppressed by at least 50 dB in the transmit pulses, meaning that almost all the second-harmonics energy captured in the echoes was produced by the reflection effect. This requires the duty cycle of the transmit pulse to be shorter than ±0.2% of a perfect 50% duty cycle.
Alternatively, in the “pulse inversion method,” inverted pulses generate pairs of phase-inverted receive signals, which are summed in the receiver to recover the second-harmonic echoes. One challenge in implementing pulse inversion is in matching the rise and fall times of the two high-voltage pulsers.
Components in each receive channel comprise a transmit/receive (T/R) switch, a low-noise amplifier (LNA), a variable-gain amplifier (VGA), an anti-alias filter (AAF), and an ADC. The T/R switch protects the LNA from the transmit pulse. Its primary design criterion is fast recovery time. The function of the LNA is obvious.
The VGA provides the dynamic range needed to accommodate signal attenuation as pulses penetrate more and more deeply into the body. Just below the patient’s skin, the return signal can be as high as 0.5 V p-p, but this signal quickly decays. Round-trip attenuation is approximately 1.4 dB/cm-MHz, and a VGA with approximately 30 to 40 dB of gain is necessary. Amplifier gain is ramped over time to match the input range of the ADC, and the results are mapped in the digital domain to accommodate the time-ramped analog gain.
The requirements for the AAF are familiar to any analog/mixed-signal designer, and the precision and signal-to-noise ratio (SNR) of the ADC are critical. Once the reflected signal is in the digital domain, the signal processing gets even more interesting. (See the Maxim Integrated Products Medical Solutions Guide at www.maxim-ic.com/landing/?lpk=517.)
In contrast to the imaging function of pulsed ultrasound, continuous-wave Doppler (CWD) is used to quantitatively measure blood flow. In CWD mode, half the ultrasound transducer elements in the array are used as phased-array transmitters. The other half are used in a receiver phased array. The CW signal frequency is typically in the range of 1 to 7.5 MHz. Achieving the lowest possible jitter in the signal source is critical.
One design challenge is that the system is looking for the small Doppler-shifted signals caused by blood flow in the presence of large stationary return signals reflected by bone and organ tissue. There are two ways to deal with this.
Many high-performance ultrasound systems simultaneously demodulate the signals from each receive channel into in-phase and quadrature (I and Q) components at the LNA output. The local-oscillator (LO) phase is adjusted to do the beam-forming. The output of all the channels is summed, band-pass filtered, digitized, and processed in the digital domain. The alternative method used in low-cost systems uses delay lines to accomplish beamforming ahead of the mixer.
In developed countries, diabetes, naturally occurring and lifestyle-driven, is widespread. Self-treatment requires patients to know their blood-glucose levels and to inject their own insulin, both of which are getting easier.
In the latest blood glucose meters, the volume requirement for blood samples is now small enough that patients can obtain their samples from less pain-sensitive sites than their fingertips. At the same time, improved test strips, electronics, and measurement algorithms have increased the precision and accuracy of measurements.
One new variety of blood glucose meter provides for “continuous” testing, instead of requiring patients to prick themselves for each test. Sometimes test results are used to control insulin pumps. Today, the control is open-loop, but it’s only a matter of time before the glucose monitor and pump constitute a closed-loop system.
Continuous meters available only by prescription employ a subcutaneous electrochemical sensor that takes measurements at fixed intervals, so they aren’t literally “continuous.” The more familiar “single-test” meters operate based on electrochemical or optical reflectometry. (Most are electrochemical.)
In electrochemical meters, a bias voltage generated by a DAC is applied across the test strip, and the current between the test probes is measured. The resistance between the test points is a function of the blood sugar in the sample. The current is usually converted to a voltage by a transimpedance amplifier and digitized by an ADC. Current levels range from 10 to 50 μA and can be resolved to less than 10 nA. The meters calibrate themselves to account for variations in ambient temperature.
Depending on the price of the meter and the test strips it uses, some calibration may be required before readings are taken. When calibration is required, it may involve manually entering a code from the test-strip package. If the patient cannot do that reliably, there are strips that provide their own calibration data via an electronic memory device. Top-of-the line units offer “self-calibration” either based on tight manufacturing controls on the strips or actual self-calibration of a reference strip included in each package of strips, with the strips being dispensed by the meter as needed.
Meanwhile, insulin pumps make diabetes management less regimented and more of a background task while providing tighter control of blood glucose than periodic injections along with greater adaptability to eating, resting, and exercising.
Insulin is administered in 0.01-ml “units” by an electrical screw-drive piston pump. Total pump cartridge volumes are 200 to 300 units. To drive the screw, pump manufacturers may use either optical encoders and dc motors, or stepper motors. Future designs may use microelectromechanical systems (MEMS) technology.
Instrumentation for flow sensing is based on strain-gauge pressure sensors with output levels in the millivolt range, so the semiconductor content is the typical AFE and ADC. To drive the pump actuator, manufacturers typically use a boost regulator to step up the battery voltage. Typical pumps operate for three to 10 weeks at a time on a single AA or AAA alkaline or lithium battery.
Digital medical thermometers respond faster and are easier to read than mercury-bulb thermometers. The ear type of thermometer remotely measures the infrared energy radiated from the ear canal. Temple and forehead types operate similarly, but have to be in contact with the body.
Probe-type contact-thermometers are usually thermistor-based. A precision resistor in series with the thermistor forms a voltage divider that is driven by a precision reference voltage (VRef). An ADC measures the drop across the thermistor. Usually, VRef for the ADC and the driving voltage of the voltage divider come from the same source, and 12 bits of ADC resolution provide sufficient precision.
For rapid response, the conversion from the ADC output to a digital reading uses a lookup table and linear interpolation. For most purposes, this is sufficient, because the range of possible temperatures in a living body is fairly narrow. Your friendly neighborhood police crime-scene investigator may face a wider range of temperatures, in which case additional calibration is welcome.
The sensors used for infrared measuring ear-type and forehead-type thermometers use thermopiles, which are series arrays of thermocouples that generate an output voltage proportion to the energy they sense. This approach requires a separate thermistor to measure the temperature at the cold-junction end of the thermopile to calculate the delta-T across the thermopile. A temporal thermometer achieves higher precision by measuring the temperature of the temporal artery, but the process is similar.