Teardown: Inside the art of pulse oximetry

The company's previous-generation Radical pulse oximeter, based on Masimo SET (signal extraction technology) two-LED technology, could extract basic information such as arterial oxygen saturation, pulse rate and perfusion (pulse strength). The Radical-7 uses the company's Rainbow SET seven-LED technology to enable the extraction of more advanced parameters, such as carboxyhemoglobin (to test for carbon monoxide poisoning), methemoglobin (to check for overdoses of painkillers such as benzocaine and for poisoning from chemicals such as nitrates) and total hemoglobin, as well as oxygen content and arterial oxygen saturation. It also charts the pleth waveform (heartbeat) and provides a pleth variability index.

The basic measurements and raw data are generated using well-trodden spectrographic principles of absorption at different wavelengths, in this case using LEDs as the light source. "We won't discuss the specific wavelengths," said Sampath, "but the two-LED [Masimo SET] system has a visible LED and a near-IR [infrared] LED, whereas the Rainbow [SET] has several visible LEDs and several near-IR LEDs."

While the use of LEDs in such applications is not new, Masimo has advanced the technology with its own design.

The real innovation, however, is the company's signal extraction technology.

At the time of its introduction, SET was a new and fundamentally distinct method of acquiring, processing and reporting arterial oxygen saturation and pulse rate data. It uses adaptive filters (widely deployed in communications systems) to accurately establish a "noise reference" in the detected physiological signal sent back from the sensor. The reference allows the direct calculation of arterial oxygen saturation and pulse rate. Because it is not bound by a conventional "red over infrared" ratio approach, the Masimo SET system greatly mitigates the problems of motion artifacts, low peripheral perfusion and most low signal-to-noise situations, allowing it to be used in low-signal, high-motion applications.

How does one derive such a reference where none is present? "This is the fundamental problem we solve with an algorithm called the Discrete Saturation Transform engine, or DST engine," said Sampath.

All algorithms are based on assumptions. The higher the number of assumptions, the weaker the algorithm. The DST engine makes only one assumption, that arterial blood has higher oxygenation than that of venous blood.

Working from that single assumption, the DST engine allows the Masimo system to separate and consequently calculate the optical density ratios that correspond to the arterial oxygen saturation and estimated venous oxygen saturation. Those optical densities are not known beforehand but are required to obtain the appropriate reference signals for adaptive noise cancellation.

The signal processing at the heart of Masimo SET proved so powerful under patient motion conditions that it became central to a patent-infringement dispute between Masimo and longtime competitor Nellcor. The issue was finally put to rest in January 2006 when, after Masimo won the patent suit in both district and appellate courts, the companies arrived at a settlement agreement.

Winning patent battles is one thing, but the proof of a good algorithm is in its efficacy—in this case, how accurate it is relative to established techniques.

According to Sampath, in carboxyhemoglobin measurements, the Masimo system achieves accuracy of +/-3 percent, vs. +/-2 percent in the lab. For methemoglobin measurements, the window narrows to +/-1 percent, "the same as with invasive techniques," he said.

For hemoglobin measurement accuracy, the figure for the Radical-7 is 1 gram per deciliter (g/dL), vs. 0.5 g/dL in the lab. "All of these are for 1 standard deviation, or 67 percent of the population," Sampath said.

Given the parameters of the design, with so much dependent upon accurate sensing of low-level signals and their back-end processing, the designers of the Radical-7 had their work cut out for them. They had to make wise design and component choices with respect to signal acquisition, conditioning, amplification, conversion and, ultimately, a signal-processing platform.