Architectural Trends in Data Acquisition Systems for Medical Applications

The role of higherintegration of both channel count and analog function can be significant inaddressing the demands in analog signal acquisition for a variety of today'smedical applications. To effectively leverage the benefits of higherintegration and still achieve a desired cost/performance target requires anunderstanding of both process advantages and design trade-offs most relevant tothe system of interest. This discussion presents a targeted comparison in thearchitectural trends and design trade-offs as it relates to improved systemperformance. We'll also cover lower cost-per-channel for analog dataacquisition systems of two very specific arenas in the medical field: ECG andCT scanning.

Click hereto download TI's Medical Applications Guides

Electrocardiography

Electrocardiography (ECG) is a noninvasive method forcapturing and processing the electrical signature of the heart via skinelectrodes. ECG applications are broad and varied in their scope, andtherefore, so are the requirements for the analog data acquisition system. TheECG signal and its harmonics are low in bandwidth (150Hz). The challenges inECG signal acquisition revolve primarily around external noise rejection on theanalog front-end (AFE), noise filtering, sampling, and signal processing by theback-end conditioning circuitry, analog-to-digital converter (ADC) andmicrocontroller unit (MCU).

A typical low-end ECG dataacquisition system requires a minimum of two to three electrodes that aresensed by a differential, analog front-end (AFE) gain block, a band pass filterand analog gain block, and a lower-resolution (10-12 bit) successiveapproximation register (SAR) converter. The strategy and degree to which noiseis removed from an ECG signal acquisition system is dependent on the overallsystem cost target. The cost of non-clinical, lower-end ECG systems has beendramatically reduced by analog integration. In fact, a disposable ECG patchmost likely will contain the AFE, filtering, gain amps and a low-resolution ADCintegrated into the MCU. Figure 1 shows a higher-level block diagram of atypical ECG data acquisition system. Note the boxed area showing the analogcontent that typically is fair game for integrating lower-end ECG dataacquisition systems.

Broader integration is alsothe trend for medium to higher end, higher lead count ECG systems that requirelower noise AFEs and more bits of resolution in the data acquisition system, aswell as integrated ECG-specific functions such as right leg drive bias, Wilsoncentral reference (for chest lead measurements), lead off detection, andprovisions to separately process pace signals. One good example of this type ofintegration is demonstrated by the ADS1298, a low-power, low-noise,simultaneous sampling, 24-bit ADC that includes all of these ECG functionsalong with a front-end multiplexor that allows the user to toggle between theelectrode inputs, input test signals, right leg drive reference, supply voltageand an internal temperature sensor.

Computed Tomography Scans

Medical computed tomography (CT) is a method used toprocess two dimensional X-rays of images ("slices") that result in athree-dimensional representation of a target area of the body. Capturing a"slice" of data is initiated when X-rays transmitted through the body are scintillatedand strike a dense photodiode array, resulting in a photocurrent that issensed, amplified, sampled and filtered by a data acquisition system comprisedof an integrating AFE, ADC, and a significant amount of post processing to takethe numerous slices of 2-D X-ray data and reconstruct the desired image.

Unlike ECG, CT scanning applications are largely clinical and aresubject to a very specific set of cost/performance trade-offs. A better imagecan be achieved with better signal-to-noise ratio (SNR). Better SNR comes byincreasing the amount of signal more than the noise. More signal can beobtained by increasing the number of photodiode detectors. Therefore, three ofthe key design constraints in CT scanning signal acquisition are SNR integralnonlinearity (INL) and channel density (i.e., area in mm2 / # channels). Sincesurface densities of state-of-the art photodiode sensor arrays in modern CTscanners have dipped below the 1mm2/channel threshold, a similar reduction inthe per channel surface density of the data acquisitionsystem translates to an increase in the amount of image data (i.e., signal)captured per slice. As a result, this increase in density often means that thephotodiode and an "integrating" AFE (see Figure 3) can be located within acloser proximity to each other, which also means that the parasitic capacitance(Cp) of the connection from the photodiode to the AFE can be further minimized.Since the voltage noise of an integrating front-end photodiode is a function ofthe ratio between the total capacitance seen at the inverting node of theintegrator (C_in), minimizing this capacitance improves the overall SNR of thedata acquisition system.

One competing effect of theincreased channel density in the integrated data acquisition system of a CTscanner is the drift of the offset and AFE transfer function integralnonlinearity due to internal self-heating. The CT data slices representsnapshots in time that rely on the absolute accuracy of the data acquisitionsystem based on an initial system calibration to reconstruct a clean 3-D imageof the desired object. For this reason, its internal drift must be kept at aminimum through low-power design techniques, packaging and component layout.Likewise, because of the "optimized" proximity of the data acquisition systemto the photodiode array, any self-heating effects may also induce self-heatingin the photodiode, which can dramatically impact the responsivity of thephotodiode and the overall SNR of the data acquisition system.

While the comparison of ECG and CT scanning spans a wide range offunctional complexity and design challenges, both are driven to reduce costwithout sacrificing performance in next-generation designs. The role of modernprocess technology and functional integration can be significant in achievingthis goal, whether it is done by integrating the AFE + ADC on the chip orincreased channel count. For the engineer to truly realize the cost/performancebenefits of advances in integration requires a solid understanding of theengineering design challenges most important to his or her design.

Matthew William Hann is Precision Analog Applicationsmanager at Texas Instruments.

For more information:
www.ti.com/medical-ca

View system block diagrams:
a euro cent ECG system www.ti.com/ecgsbd-ca
a euro cent CT scanner www.ti.com/ctsbd-ca