If you have bone shapes before the heat traeatment you can use original
images as "target" and depict warped zones and deformation patterns by thin
plate splines. I m not sure, but if you hope that observed deformation is
similar between the same kind of bone, you could build a regression model
with shape as dependent variable and (for example) heat levels as the
predictor with some covariables like bone density. But or course, you need
to register the same bone shape while exposed it to diffetent
temperatures.
best
M
2014-12-26 15:35 GMT-03:00 J Coelho <[email protected]>:
> Greetings Morphmet members,
>
> Getting to the point: I have access to an osteological collection, where
> bones are being experimentally heated at a range of different temperatures
> and times. Starting January, I'll be 3D-scanning these previous and after
> to the event. As you can imagine, bones get severely deformed by suffering
> bone warping, shrinkage, fractures and colour alterations. I'm mostly
> interested in shape modifications such as the warping and that's why I'll
> be using GMM, but I'd like to be able to do something akin to "reverse
> engineering", in order to transform burnt bones into the respective
> original form. From what I understood TPS can do this quite easily.
> However, can it help me create a model to predict the original form of
> burnt bones that didn't make part of the original analysis (akin to the
> idea of a testing/validation set in machine learning)?
>
> This is because there are already 10 burnt individuals in the collection,
> done before I talked to the main researcher about the potential of using
> GMM and 3D-Scanning, and since a lot of measurements and data where
> registered previously to the thermal alterations in these 10 individuals, I
> thought they would be ideal to test a model created through the sample of
> the next individuals that will become part of the experiment.
>
> I couldn't find any design experiments using GMM similar to this in the
> literature. So I'm a bit lost in how to proceed in the later steps, and I
> sincerely don't want things to go wrong. So if anyone could recommend
> statistical techniques I should use or test in order to create a predictive
> model to recreate original form I would be very grateful.
>
> P.S. Also, worth of mention: We have considerable data for each
> individual, with variables such as age-at-death, sex, bone weight, etc.
>
> --
> MORPHMET may be accessed via its webpage at http://www.morphometrics.org
>
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
>
--
Dr.Marcelo Cardillo
IMHICIHU-CONICET-UBA
Saavedra 15, 5to piso cp (1083)
Teléfono: (011) 4953-8548 - Interno 212
Buenos Aires. Capital Federal. Argentina
http://conicet.academia.edu/MarceloCardillo
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
