1 Looking at microCT data of Brassica pods
2 ========================================
4 I am not a biologist, please stop me and correct me if I say silly things.
24 .. image:: brassica_pod_lq.png
27 * Image analysis produces many false seeds at the beak tip
28 * Density and size is comparable to seed
29 * Hard to recognise by graphical methods alone
30 * Recognise them by mathematical means instead
35 * For every CT slice we have the centroid of the object
36 * Fit X and Y position as cubic functions of z
37 * Define 'real z' as the distance measured along the fitted curve from
38 the beak to the z coordinate of the point
40 .. image:: fitting_debug.svg
41 :target: fitting_debug.gif
45 Distinguishing between beak tip and Real Seeds™
46 -----------------------------------------------
50 1. Assert that seeds must not be implausible - Removed insufficiently many seeds
52 * Too close to the ends of the pod
53 * Too large given pod dimensions
55 2. Real z position of seeds of a pod is a sample from some probability
56 distribution, fit and paramterize the distribution to classify seeds.
58 * Sum of two normal(-ish) distributions - noise at beak might be normal,
59 everything else definitely is not
60 * More complicated distribution - too complicated
62 3. K-Means clustering - Silly for 1 dimensional data
63 4. Jenks Natural Breaks Optimisation - Should work in theory, did not work well
66 Break at Minimum Kernel Density Estimation (KDE)
67 ------------------------------------------------
69 * Beak has no Real Seeds™ and low density
70 * Expect a gap in real z of detected seeds
72 .. image:: plot_real_zs_genotype_unfiltered.png
78 * Use KDE to find density of seeds as function of real z
80 .. image:: kde_debug.png
83 * First seed has real z less than 100?
84 * Find the local minimum at lowest real z where log(KDE)<-10
85 * Keep seeds with greater real z
88 .. image:: plot_real_zs_genotype_filtered.png
91 .. image:: plot_real_zs_genotype_unfiltered.png
94 Beak and Silique length
95 -----------------------
97 Use the seed with lowest real z to mark the boundary of beak and silique:
99 .. image:: plot_bar_silique_lengths.png
100 .. image:: plot_bar_beak_lengths.png