I have curves of the following structure (it is the blue one I am interested in)
These curves reflect the volume of blood (actually gamma ray counts) in the left ventricle as a function of time (frame number). Data point values for the given curve is attached.
In the analysis of these curves (qualitative) an experienced reporter will grade the first segment (emptying) of the curve (from frame 1 to minimum) as normal, slightly delayed, delay or markedly delayed. This is based on impression of time to minimum and looking at the gradient (i.e. very arbitrary). The second (filling) portion of the curve (minimum to end frame) is also graded using the same scoring system (1-4).
Is it possible to use machine learning in this situation to train the computer to categorise the emptying and filling parts of the curve. i.e. could I use a training set of curves with known emptying and filling scores and then apply this 'knowledge' to new curves. I would like to see how well a machine and experienced reporter agree. I predict not very well as I feel the reporting system is quite arbitrary.
Absolute counts in each frame might vary by an order of magnitude. It will be the gradient and relative time to minima which will affect scoring
Is machine learning a sensible way to tackle this? How do I organise the data and tackle the analysis?
I would use R for this
Frame,Count,Curve 1,9083.06,ED_TAC 2,8998.62,ED_TAC 3,8750.27,ED_TAC 4,8310.03,ED_TAC 5,7721.91,ED_TAC 6,7055.92,ED_TAC 7,6409.06,ED_TAC 8,5864.33,ED_TAC 9,5444.69,ED_TAC 10,5181.14,ED_TAC 11,5150.63,ED_TAC 12,5331.14,ED_TAC 13,5630.62,ED_TAC 14,6085.05,ED_TAC 15,6677.4,ED_TAC 16,7312.65,ED_TAC 17,7820.77,ED_TAC 18,8111.76,ED_TAC 19,8245.61,ED_TAC 20,8348.35,ED_TAC 21,8484.98,ED_TAC 22,8678.54,ED_TAC 23,8800.05,ED_TAC 24,8752.54,ED_TAC