plt.savefig(os.path.join(out_dir_noAdapt, "%03.0d_hist.png"%i)) plt.close() def cut_off_values_upper_lower_percentile(image, percentile_lower=0.2, percentile_upper=99.8): cut_off_lower = np.percentile(image[image!=image[0,0,0]].ravel(), percentile_lower) cut_off_upper = np.percentile(image[image!=image[0,0,0]].ravel(), percentile_upper) image[image < cut_off_lower] = cut_off_lower image[image > cut_off_upper] = cut_off_upper return image percentile_target = [0., 100.] for i in range(999): print i t1_img, t1km_img, flair_img, adc_img, cbv_img, _, seg_combined = load_patient_with_t1(i, [1., 0.5, 0.5], 'new') if t1_img is not None: np.save(os.path.join(out_dir_resampled, "%03.0d_t1.npy" % i), t1_img) np.save(os.path.join(out_dir_resampled, "%03.0d_t1km.npy" % i), t1km_img) np.save(os.path.join(out_dir_resampled, "%03.0d_flair.npy" % i), flair_img) np.save(os.path.join(out_dir_resampled, "%03.0d_adc.npy" % i), adc_img) np.save(os.path.join(out_dir_resampled, "%03.0d_cbv.npy" % i), cbv_img) np.save(os.path.join(out_dir_resampled, "%03.0d_seg.npy" % i), seg_combined) for i in range(119, 150): if not os.path.isfile(os.path.join(out_dir_resampled, "%03.0d_t1.npy" % i)): continue t1_img = np.load(os.path.join(out_dir_resampled, "%03.0d_t1.npy" % i)) t1km_img = np.load(os.path.join(out_dir_resampled, "%03.0d_t1km.npy" % i))
all_min_vals_t1c = [] all_min_vals_flair = [] all_min_vals_adc = [] all_min_vals_cbv = [] all_std_t1 = [] all_std_t1c = [] all_std_flair = [] all_std_adc = [] all_std_cbv = [] all_shapes = [] for k in training_patients: print k t1_img, t1c_img, flair_img, adc_img, cbv_img, t1km_downsampled, seg = load_patient_with_t1(k, [1,1,1], 'new') if t1_img is None: continue pool = Pool(5) t1_img, t1c_img, flair_img, adc_img, cbv_img = pool.map(cut_off_values_upper_lower_percentile, (t1_img, t1c_img, flair_img, adc_img, cbv_img)) pool.close() pool.join() brain_mask = seg!=0 all_means_t1.append(t1_img[brain_mask].mean()) all_means_t1c.append(t1c_img[brain_mask].mean()) all_means_flair.append(flair_img[brain_mask].mean()) all_means_adc.append(adc_img[brain_mask].mean()) all_means_cbv.append(cbv_img[brain_mask].mean())
percentile_lower=0.2, percentile_upper=99.8): cut_off_lower = np.percentile(image[image != image[0, 0, 0]].ravel(), percentile_lower) cut_off_upper = np.percentile(image[image != image[0, 0, 0]].ravel(), percentile_upper) image[image < cut_off_lower] = cut_off_lower image[image > cut_off_upper] = cut_off_upper return image percentile_target = [0., 100.] for i in range(999): print i t1_img, t1km_img, flair_img, adc_img, cbv_img, _, seg_combined = load_patient_with_t1( i, [1., 0.5, 0.5], 'new') if t1_img is not None: np.save(os.path.join(out_dir_resampled, "%03.0d_t1.npy" % i), t1_img) np.save(os.path.join(out_dir_resampled, "%03.0d_t1km.npy" % i), t1km_img) np.save(os.path.join(out_dir_resampled, "%03.0d_flair.npy" % i), flair_img) np.save(os.path.join(out_dir_resampled, "%03.0d_adc.npy" % i), adc_img) np.save(os.path.join(out_dir_resampled, "%03.0d_cbv.npy" % i), cbv_img) np.save(os.path.join(out_dir_resampled, "%03.0d_seg.npy" % i), seg_combined) for i in range(119, 150): if not os.path.isfile(os.path.join(out_dir_resampled, "%03.0d_t1.npy" % i)): continue