Dr N R Shanker, a researcher at Sree Sastha Institute of Engineering and Technology, India has developed an artificial neural network that can analyze the photo of an hand and determine the cholesterol content. The technique that was used was standard neural network analysis. Create a training set with pictures of the hand and cholesterol values. Use that to train a neural network such that it can classify a pattern that is predictive. The method they used was the mean algorithm and used that to form a correlation table.
Interestingly, the question to ask is whether the results mean anything significant and can some biological meaning be drawn from that? One hypothesis is that the cholesterol deposits near the skin surface and may lead to a characteristic set of wrinkles or indentations. These would be impossible to distinguish by a human but given a large enough data set, it might be possible to understand patterns in the images.
However, the dangers are many. This article was published in a peer reviewed Journal so there must have been a detailed review process – Int. J. of Medical Engineering and Informatics, 2012 Vol.4, No.3, pp.223 – 230. The one that is a always considered a problem in these correlations is the danger of over-fitting. Obviously, a correlation was found but then was the correlation so fitted to the data set that if there are other hand anomalies, the correct diagnosis gets missed? Unknown but the results are very interesting for image analysis as well as heart disease since cholesterol has been the only workable correlate between cardiovascular disease and a blood “biomarker”.