« Elegant Electronics | Main | Carnival of Tomorrow #17 »

Cancer Sniffing Dog Update

dog finds cancerBack in 2004 we reported on a UK study that showed dogs were able to positively detect bladder and kidney cancer from urine samples.

One sample that was thought to be disease-free kept testing positive with the dogs. The researchers went back and reexamined the volunteer. The volunteer had kidney cancer.

But their success rate was only 41%. In a new study in San Alselmo, Californian, researchers got better results testing different forms of cancer.

The dogs correctly detected 99% of the lung cancer samples, and made a mistake with only 1% of the healthy controls. With breast cancer, they correctly detected 88% of the positive samples, and made a mistake on only 2% of the controls.

These trained dogs do as well as the current state-of-the-art testing. Perhaps false-positives and false-negatives could be reduced by employing both dogs and standard lab tests.

If a lab doesn't want to keep actual dogs on hand, perhaps they could employ dog-on-a-chip technology to get similar results.


Here's a puzzlement. Yes, probably the medical industry will go with dog-on-a-chip sensing. The question is why. I seriously doubt the cost of acquisition, operator time, and maintenance on this artificial nose will be lower than the cost of maintaining a dog and paying a trained handler. Other than cold wet nose in sensitive places, what's the big problem with using actual four legged medical workers?

We, particularly Americans, seem to have a problem with employing our animal friends in serious work, despite their proven records sniffing drugs, explosives, the living and the dead; guarding things, and so forth. Any thoughts as to why?



That dog-on-a-chip was originally being developed to allow every cop to have the benefit of a drug-sniffing dog in a PDA sized unit.

I wouldn't think that portability would be as critical in a lab as with a cop on the street. The cost of running a kennel would have to be weighed against the cost and relative benefit of dog-on-a-chip.

One thing is for sure, no matter how much you praised it, the dog-on-a-chip would not be as friendly. :-)

Mike brought up the same arguments in his email reply to me. Let me cut and paste his objections to a dog first.

[...]I remember encountering the LTCO figures for a household pet some time ago (this was a study commissioned by one of the animal food companies, I believe) that placed the total investment in a pet (four-legged predetory mammal of house cat size or greater) at approximately the same cost as raising a child through high school.

Add to that cost:

Training: $ Thousand(s)

Handler: $ Thousands per year, may be spread across multiple “sensor platforms”

Handler’s Training: $ Thousand(s)

Also, a bench-top box (if it can be constructed) doesn’t pose a liability risk of destruction or, God forbid, attack, and you begin to see where tech may be more cost-effective.

Finally, there is the issue of communications. Between human and canine there will, at least for the near future, always be a degree of ambiiguity. A technical solution might only have GO / NO GO lights.[...]

I think perhaps we have a fundamental difference in expectation of how this tool would be used. I see the nose-in-a-box as yet another expensive tool that will only be available in labs and at great expense. Even if the technology is cheap by nature, the process of FDA certification, liability insurance, operator training, certified maintenance, and so forth will raise the cost far beyond any reasonable point.
By contrast, I see the dog as your GP's pet. At the end of the day your cancer scanner goes home with its doctor owner, so your maintenance cost becomes less relivant. The doctor is trained to work with the dog, or more likely trained with the dog in the first place, so you need not maintain a separate technical staff. Dog maintenance systems are already established and very cost effective. (Hell, my vet's so good, I envy my cat her quality of health care.) Dog firmware is well established and has a hundred-thousand year reputation for robustness. And even if the dog does flake, even if the dog bites thet patient or just gets bored and stops paying attention, this sort of failure will be obvious, whereas software bugs, sensor failures, operator errors, and so forth may occur silently.

Post a comment