Food scanners and chemometric calibration
Submitted by Robin on 22 October 2015 - 9:22am
Dear all,
I would like to start a discussion about so-called food scanners, which are currently being advertised by TellSpec (www.tellspec.com) and SCiO (www.consumerphysics.com and www.dietsensor.com) and the technology behind them. These mobile devices employ near-infrared spectroscopy to determine the chemical makeup of foods and should be able to predict the amount of calories, fats, carbohydrates and proteins. In my opinion the main challenge here lies not the miniaturization of near-infrared spectroscopy, but in the development of appropriate chemometric models. The calibration models must be able to quantify the concentrations of ingredients over a vast amount of highly variable food products and even for unknown and homemade foods. From my understanding, traditional chemometric calibration models (linear and non-linear) cannot be applied here, since these are highly problem and food specific. So what is needed here is a kind of “universal” calibration approach that is independent of the sample under investigation. What do you think: Is it possible to build such a universal calibration model? If so, how can this be done?
Regards,
Robin
ianm
22 Oct 2015
Robin, I am at NIR-2015 this week. Yesterday in his plenary on handheld instruments, Heinz Siesler mentioned these. If I remember correctly, he said that since both did not publish any technical detail, it was difficult to judge their capabilities. Of course, they are also not available yet. I will and get Heinz and some others to comment.
Ian
Permalink
Robin
22 Oct 2015
Hi Ian, thanks a lot, that would be great!
Permalink
rcmartins
22 Oct 2015
As far as there is public information is contained in these patents:
https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2013065035
and
http://www.google.com/patents/US20150036138
RCM
Permalink
Robin
24 Oct 2015
Thanks, RCM. Indeed, the TellSpec patent gives some insight into the chemometrics.
Permalink