Mercurial > repos > greg > linear_fascile_evaluation

comparison linear_fascile_evaluation.py @ 7:eb03934e044f draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author greg
date Wed, 29 Nov 2017 16:40:08 -0500
parents 0ddfcb3b5ce6
children 2de70534993d
comparison
equal deleted inserted replaced
6:8dba8c7c1f53 7:eb03934e044f
18 import numpy as np 18 import numpy as np
19 19
20 parser = argparse.ArgumentParser() 20 parser = argparse.ArgumentParser()
21 parser.add_argument('--input', dest='input', help='Track Visualization Header dataset') 21 parser.add_argument('--input', dest='input', help='Track Visualization Header dataset')
22 parser.add_argument('--output_life_candidates', dest='output_life_candidates', help='Output life candidates') 22 parser.add_argument('--output_life_candidates', dest='output_life_candidates', help='Output life candidates')
23 parser.add_argument('--output_life_optimized', dest='output_life_optimized', help='Output life optimized streamlines')
24 parser.add_argument('--output_beta_histogram', dest='output_beta_histogram', help='Output beta histogram')
25 parser.add_argument('--output_error_histograms', dest='output_error_histograms', help='Output error histograms')
26 parser.add_argument('--output_spatial_errors', dest='output_spatial_errors', help='Output spatial errors')
27 23
28 args = parser.parse_args() 24 args = parser.parse_args()
29 25
30 # We'll need to know where the corpus callosum is from these variables. 26 # We'll need to know where the corpus callosum is from these variables.
31 hardi_img, gtab, labels_img = read_stanford_labels() 27 hardi_img, gtab, labels_img = read_stanford_labels()
52 fvtk.add(ren, cc_ROI_actor) 48 fvtk.add(ren, cc_ROI_actor)
53 fvtk.add(ren, vol_actor) 49 fvtk.add(ren, vol_actor)
54 fvtk.add(ren, vol_actor2) 50 fvtk.add(ren, vol_actor2)
55 fvtk.record(ren, n_frames=1, out_path="life_candidates.png", size=(800, 800)) 51 fvtk.record(ren, n_frames=1, out_path="life_candidates.png", size=(800, 800))
56 shutil.move("life_candidates.png", args.output_life_candidates) 52 shutil.move("life_candidates.png", args.output_life_candidates)
57 # Initialize a LiFE model.
58 fiber_model = life.FiberModel(gtab)
59 # Fit the model, producing a FiberFit class instance,
60 # that stores the data, as well as the results of the
61 # fitting procedure.
62 fiber_fit = fiber_model.fit(data, candidate_sl, affine=np.eye(4))
63 fig, ax = plt.subplots(1)
64 ax.hist(fiber_fit.beta, bins=100, histtype='step')
65 ax.set_xlabel('Fiber weights')
66 ax.set_ylabel('# fibers')
67 fig.savefig("beta_histogram.png")
68 shutil.move("beta_histogram.png", args.output_beta_histogram)
69 # Filter out these redundant streamlines and
70 # generate an optimized group of streamlines.
71 optimized_sl = list(np.array(candidate_sl)[np.where(fiber_fit.beta > 0)[0]])
72 ren = fvtk.ren()
73 fvtk.add(ren, fvtk.streamtube(optimized_sl, line_colors(optimized_sl)))
74 fvtk.add(ren, cc_ROI_actor)
75 fvtk.add(ren, vol_actor)
76 fvtk.record(ren, n_frames=1, out_path="optimized.png", size=(800, 800))
77 shutil.move("optimized.png", args.output_life_optimized)
78 model_predict = fiber_fit.predict()
79 # Focus on the error in prediction of the diffusion-weighted
80 # data, and calculate the root of the mean squared error.
81 model_error = model_predict - fiber_fit.data
82 model_rmse = np.sqrt(np.mean(model_error[:, 10:] ** 2, -1))
83 # Calculate another error term by assuming that the weight for each streamline
84 # is equal to zero. This produces the naive prediction of the mean of the
85 # signal in each voxel.
86 beta_baseline = np.zeros(fiber_fit.beta.shape[0])
87 pred_weighted = np.reshape(opt.spdot(fiber_fit.life_matrix, beta_baseline), (fiber_fit.vox_coords.shape[0], np.sum(~gtab.b0s_mask)))
88 mean_pred = np.empty((fiber_fit.vox_coords.shape[0], gtab.bvals.shape[0]))
89 S0 = fiber_fit.b0_signal
90 # Since the fitting is done in the demeaned S/S0 domain,
91 # add back the mean and then multiply by S0 in every voxel:
92 mean_pred[..., gtab.b0s_mask] = S0[:, None]
93 mean_pred[..., ~gtab.b0s_mask] = (pred_weighted + fiber_fit.mean_signal[:, None]) * S0[:, None]
94 mean_error = mean_pred - fiber_fit.data
95 mean_rmse = np.sqrt(np.mean(mean_error ** 2, -1))
96 # Compare the overall distribution of errors
97 # between these two alternative models of the ROI.
98 fig, ax = plt.subplots(1)
99 ax.hist(mean_rmse - model_rmse, bins=100, histtype='step')
100 ax.text(0.2, 0.9, 'Median RMSE, mean model: %.2f' % np.median(mean_rmse), horizontalalignment='left', verticalalignment='center', transform=ax.transAxes)
101 ax.text(0.2, 0.8, 'Median RMSE, LiFE: %.2f' % np.median(model_rmse), horizontalalignment='left', verticalalignment='center', transform=ax.transAxes)
102 ax.set_xlabel('RMS Error')
103 ax.set_ylabel('# voxels')
104 fig.savefig("error_histograms.png")
105 shutil.move("error_histograms.png", args.output_error_histograms)
106 # Show the spatial distribution of the two error terms,
107 # and of the improvement with the model fit:
108 vol_model = np.ones(data.shape[:3]) * np.nan
109 vol_model[fiber_fit.vox_coords[:, 0], fiber_fit.vox_coords[:, 1], fiber_fit.vox_coords[:, 2]] = model_rmse
110 vol_mean = np.ones(data.shape[:3]) * np.nan
111 vol_mean[fiber_fit.vox_coords[:, 0], fiber_fit.vox_coords[:, 1], fiber_fit.vox_coords[:, 2]] = mean_rmse
112 vol_improve = np.ones(data.shape[:3]) * np.nan
113 vol_improve[fiber_fit.vox_coords[:, 0], fiber_fit.vox_coords[:, 1], fiber_fit.vox_coords[:, 2]] = mean_rmse - model_rmse
114 sl_idx = 49
115 fig = plt.figure()
116 fig.subplots_adjust(left=0.05, right=0.95)
117 ax = AxesGrid(fig, 111, nrows_ncols=(1, 3), label_mode="1", share_all=True, cbar_location="top", cbar_mode="each", cbar_size="10%", cbar_pad="5%")
118 ax[0].matshow(np.rot90(t1_data[sl_idx, :, :]), cmap=matplotlib.cm.bone)
119 im = ax[0].matshow(np.rot90(vol_model[sl_idx, :, :]), cmap=matplotlib.cm.hot)
120 ax.cbar_axes[0].colorbar(im)
121 ax[1].matshow(np.rot90(t1_data[sl_idx, :, :]), cmap=matplotlib.cm.bone)
122 im = ax[1].matshow(np.rot90(vol_mean[sl_idx, :, :]), cmap=matplotlib.cm.hot)
123 ax.cbar_axes[1].colorbar(im)
124 ax[2].matshow(np.rot90(t1_data[sl_idx, :, :]), cmap=matplotlib.cm.bone)
125 im = ax[2].matshow(np.rot90(vol_improve[sl_idx, :, :]), cmap=matplotlib.cm.RdBu)
126 ax.cbar_axes[2].colorbar(im)
127 for lax in ax:
128 lax.set_xticks([])
129 lax.set_yticks([])
130 fig.savefig("spatial_errors.png")
131 shutil.move("spatial_errors.png", args.output_spatial_errors)