What the Nose Knows - Avery Gilbert [107]
Like the Freudians, futurists are quick to dismiss the sense of smell as an evolutionary dead end. They speculate that our noses will shrink and our smelling ability will devolve along with it. But is this really our fate? To peer into our olfactory future, we must look toward smelling machines and olfactory genes.
Unlike space aliens, electronic noses are already among us—the first commercial units were delivered around 1992, intended for use in quality control in the flavor and fragrance industries. An e-nose uses an array of chemical sensors to detect odor molecules, and pattern analysis software to distinguish between them. Early models were large boxes that sat in the laboratory; more-recent handheld versions resemble something the meter reader might carry. What sets the e-nose apart from other chemical detectors—like those that measure breath alcohol or warn of carbon monoxide—is that it responds to a broad range of molecules. (Smoke alarms that work on optical principles are even less specific, which is why they sometimes mistake steam or fine dust for smoke.) The chemical sensors of an e-nose can be made from all sorts of materials, with conducting polymers being a popular choice. A conducting polymer changes its electrical resistance in the presence of volatile molecules. Some versions respond to odor at concentrations near the limits of human perception. These polymers are sensitive but not sophisticated; they are basically chemical sponges with different absorbent qualities.
The usefulness of an e-nose depends on its software as much as its sensors. The software extracts a pattern from the sensor input using formidable statistical methods. Multiple sensors give the e-nose a big advantage over single-molecule detectors. In particular, they avoid the pitfall of cross-interference. Imagine a fart detector that works by responding to a single molecule, namely hydrogen sulfide. Embarrassingly, it would go off every time your mom makes some egg salad. In contrast, a broadband e-nose reads the hydrogen sulfide along with other molecules, and would be less likely to mistakenly insult the lady of the house.
How well does an e-nose actually perform? Does it have the potential to take jobs away from humans? Early models were intensely hyped by their manufacturers, and when the devices failed to live up to expectations, customers were left with a lingering negative impression of the technology. The hype hasn’t entirely disappeared. An informal test in 2006 concluded that one brand of consumer e-nose—a handheld, battery-operated model that detected spoiled meat using the amines released by contaminating bacteria—oversold both the accuracy and the benefits of the device.
In general, the practical skills of the e-nose are real but modest; they include telling whether two smells are the same or different. This simple talent is useful in quality control where a manufacturer needs to keep batch-to-batch variation within limits or reject tainted raw materials. An e-nose excels at same/different judgments, and unlike human sensory panelists, it doesn’t get tired or bored. (This doesn’t mean it’s maintenance-free; e-noses have to be recalibrated frequently owing to “sensor drift.”) E-noses are good for dirty and dangerous jobs that humans don’t want, such as monitoring emissions from animal feed lots and sewage treatment plants, or searching for land mines.
The e-nose also has a future in medicine. One device can detect diabetes from volatiles in the breath of a patient; another can find evidence of lung cancer. (Those cancer-sniffing dogs might be out of work before they know it.) An e-nose diagnostic scan would be quick and noninvasive. The main technical challenge is detecting a disease-related odor signal against a varying background of body odor.
Potential consumer