Signal Processing Algorithms in Glucometers

Signal Processing Algorithms in Glucometers

A blood glucose meter never sees sugar. It sees a small electrical signal from a chemical reaction on the test strip, and its software decides whether that signal can be trusted and what number it stands for. When we read FDA decision summaries for several meters, the glucose measurement turned out to be one step in a longer sequence: confirm the strip is full, probe the sample with alternating current, measure glucose, then combine everything into a corrected result [1].

The basic signal

Most strips are amperometric, which means the meter applies a voltage and measures the current that flows. An enzyme on the strip reacts with glucose. A helper chemical called a mediator carries the freed electrons to an electrode, and the resulting current is proportional to the glucose in the sample [2]. Turning that current into mg/dL takes a calibration line (a slope and an intercept), and each manufacturing lot of strips has its own [3].

FreeStyle strips from Abbott take a different route called coulometry. Instead of reading the current at one moment, the meter adds up the total charge the reaction produces. Measurement time grows with the glucose level, and the 2008 paper describing the system says that variable timing also compensates for reaction-speed changes caused by temperature and hematocrit [3]. Our article on coulometry vs. amperometry compares the two methods in more depth.

Safety gating: checking the sample first

The first algorithm runs before any glucose reading. In the Accu-Chek Aviva Plus system, the meter applies a small alternating current (AC) to a fresh strip and waits for blood to cause a conductivity spike at the measurement electrode and a separate sample-sufficiency electrode, which together confirm that enough blood is present [1]. FreeStyle strips do the same job with two fill-indicator electrodes across the sample channel; the test doesn't start until the sample is sufficient [3].

FDA expects this gate. Its guidance for over-the-counter meters, issued September 29, 2020, says a meter should detect a short sample and not report a result, and that detection shouldn't depend on the user eyeballing the strip [4]. It also recommends that meters recognize a used strip on their own and refuse to report a result from it [4]. Some meters accept a second drop by design: FDA's November 2016 summary for the Contour Next One lists a 60-second window to reapply blood, up from 30 seconds on the older meter it was compared with [5].

AC impedance and hematocrit

Hematocrit is the share of blood volume taken up by red blood cells. It matters because meters without correction tend to read high when hematocrit is low, as in anemia, and low when it is high [6]. A person's hematocrit isn't fixed either. A 2011 study lists smoking, exercise, pregnancy, time in the mountains and hemodialysis among the things that shift it [6].

Measuring how the sample responds to alternating current, its impedance, is one way meters correct for this. After the fill check, the Accu-Chek Aviva meter, used with Aviva Plus strips, applies AC voltages at four frequencies and reads the responses, which carry information about the sample and the surrounding temperature. Only then does it apply a small direct-current (DC) voltage and read the current from the glucose reaction. The meter combines the AC and DC information into a result compensated for hematocrit and temperature, and FDA's summary reports accurate results across the claimed hematocrit range of 10 to 65% [1]. FDA's August 2016 summary for the Accu-Chek Guide says its sample and environmental conditions "are evaluated using AC and DC signals," with the same 10 to 65% range [7].

Here's what surprised us: in 2011, correction was far from universal. Researchers tested 12 self-test meters on venous blood adjusted to different hematocrit levels. Four stayed stable, and the rest deviated by 16 to 36% [6]. Three of the four stable meters used conventional "static" measurement, so stability didn't depend on any single technique [6]. The results describe meters as sold in 2011, not a ranking of today's models.

Dynamic electrochemistry and multi-pulse measurements

Dynamic electrochemistry takes several readings under different conditions, varying frequency and voltage and adjusting the input signal as the reaction progresses. The study's authors call the richer output a "fingerprint," from which the meter's algorithm derives correction factors [6].

The math treats each substance that reacts at the electrode as having its own characteristic phase angle, so comparing responses under different conditions lets the algorithm estimate how much of the signal comes from glucose and how much from known interfering substances, including ascorbic acid (vitamin C) [6]. In the same lab test, the BGStar meter, which used this method, deviated 8% across hematocrit levels [6]. That study was funded by an unrestricted grant from sanofi-aventis, the company behind the BGStar [6].

Corrections have limits, and the strip labeling is where they are listed. Despite its AC correction, the Aviva Plus labeling warns that intravenous vitamin C producing blood levels above 3 mg/dL will cause overestimated results, as will galactose above 15 mg/dL [1]. Some strips reduce interference through chemistry rather than math. FreeStyle strips run the mediator at a low voltage, which the 2008 paper says minimizes interference from acetaminophen, ascorbic acid and uric acid [3]. More on those chemistries is in hematocrit and interference correction.

Auto-calibration and no-coding meters

Older meters had to be set for each strip lot, by entering a code or inserting a code chip, so they would use the right calibration line [3]. People sometimes got it wrong. Across four small studies of about 500 patients, two found about 3% had miscoded their meters and two found about 16%. In a separate study where meters were miscoded on purpose, median errors were about 30% in either direction [8]. Manufacturers responded in three ways: strip cartridges that carry the code for the meter to read, meters fixed to the most common code, and true no-code systems with less variation between lots [8].

FreeStyle Lite is the third kind. Because coulometric charge depends on sample volume, lot-to-lot variation could be held down by controlling the size of the sample chamber, and the meter shipped preconfigured with one set of calibration values [3]. Electrode design also keeps the wrong strips out. The meter's strip port switches on only for strips carrying the FreeStyle Lite version of a "turn-on bar" electrode, so older FreeStyle strips can't be used in it [3]. In FDA's 2016 summary for the Accu-Chek Guide, the coding entry simply reads "None" [7]. Our calibration codes vs. no-code article covers the history of coding.

What the summaries leave out

Each correction is validated only over a claimed range, such as the 10 to 65% hematocrit span on the Aviva Plus and Guide systems [1] [7]. The summaries we read rarely spell out the math. FDA's Contour Next One summary says only that new meter software converts current into glucose and that a revised calculation algorithm improves accuracy over the earlier Contour Next system [5]. Whether a particular meter suits someone with anemia, someone on dialysis, or someone receiving high-dose intravenous vitamin C is a question for their care team.