def resample_image(fname, suffix='_resampled.nii.gz', binary=False, npx=0.3, npy=0.3, thr=0.0, interpolation='spline'): """ Resampling function: add a padding, resample, crop the padding :param fname: name of the image file to be resampled :param suffix: suffix added to the original fname after resampling :param binary: boolean, image is binary or not :param npx: new pixel size in the x direction :param npy: new pixel size in the y direction :param thr: if the image is binary, it will be thresholded at thr (default=0) after the resampling :param interpolation: type of interpolation used for the resampling :return: file name after resampling (or original fname if it was already in the correct resolution) """ im_in = Image(fname) orientation = im_in.orientation if orientation != 'RPI': im_in.change_orientation('RPI') fname = add_suffix(im_in.absolutepath, "_rpi") im_in.save(path=fname, mutable=True) nx, ny, nz, nt, px, py, pz, pt = im_in.dim if np.round(px, 2) != np.round(npx, 2) or np.round(py, 2) != np.round(npy, 2): name_resample = extract_fname(fname)[1] + suffix if binary: interpolation = 'nn' if nz == 1: # when data is 2d: we convert it to a 3d image in order to avoid conversion problem with 2d data # TODO: check if this above problem is still present (now that we are using nibabel instead of nipy) run_proc(['sct_image', '-i', ','.join([fname, fname]), '-concat', 'z', '-o', fname]) run_proc(['sct_resample', '-i', fname, '-mm', str(npx) + 'x' + str(npy) + 'x' + str(pz), '-o', name_resample, '-x', interpolation]) if nz == 1: # when input data was 2d: re-convert data 3d-->2d run_proc(['sct_image', '-i', name_resample, '-split', 'z']) im_split = Image(name_resample.split('.nii.gz')[0] + '_Z0000.nii.gz') im_split.save(name_resample) if binary: img = Image(name_resample) img.data = binarize(img.data, thr) img.save() if orientation != 'RPI': img = Image(name_resample) img.change_orientation(orientation) name_resample = add_suffix(img.absolutepath, "_{}".format(orientation.lower())) img.save(path=name_resample, mutable=True) return name_resample else: if orientation != 'RPI': fname = add_suffix(fname, "_RPI") im_in = change_orientation(im_in, orientation).save(fname) printv('Image resolution already ' + str(npx) + 'x' + str(npy) + 'xpz') return fname
def merge_images(list_fname_src, fname_dest, list_fname_warp, param): """ Merge multiple source images onto destination space. All images are warped to the destination space and then added. To deal with overlap during merging (e.g. one voxel in destination image is shared with two input images), the resulting voxel is divided by the sum of the partial volume of each image. For example, if src(x,y,z)=1 is mapped to dest(i,j,k) with a partial volume of 0.5 (because destination voxel is bigger), then its value after linear interpolation will be 0.5. To account for partial volume, the resulting voxel will be: dest(i,j,k) = 0.5*0.5/0.5 = 0.5. Now, if two voxels overlap in the destination space, let's say: src(x,y,z)=1 and src2'(x',y',z')=1, then the resulting value will be: dest(i,j,k) = (0.5*0.5 + 0.5*0.5) / (0.5+0.5) = 0.5. So this function acts like a weighted average operator, only in destination voxels that share multiple source voxels. Parameters ---------- list_fname_src fname_dest list_fname_warp param Returns ------- """ # create temporary folder path_tmp = tmp_create() # get dimensions of destination file nii_dest = Image(fname_dest) # initialize variables data = np.zeros([ nii_dest.dim[0], nii_dest.dim[1], nii_dest.dim[2], len(list_fname_src) ]) partial_volume = np.zeros([ nii_dest.dim[0], nii_dest.dim[1], nii_dest.dim[2], len(list_fname_src) ]) data_merge = np.zeros([nii_dest.dim[0], nii_dest.dim[1], nii_dest.dim[2]]) # loop across files i_file = 0 for fname_src in list_fname_src: # apply transformation src --> dest sct_apply_transfo.main(argv=[ '-i', fname_src, '-d', fname_dest, '-w', list_fname_warp[i_file], '-x', param.interp, '-o', 'src_' + str(i_file) + '_template.nii.gz', '-v', str(param.verbose) ]) # create binary mask from input file by assigning one to all non-null voxels img = Image(fname_src) out = img.copy() out.data = binarize(out.data, param.almost_zero) out.save(path=f"src_{i_file}native_bin.nii.gz") # apply transformation to binary mask to compute partial volume sct_apply_transfo.main(argv=[ '-i', 'src_' + str(i_file) + 'native_bin.nii.gz', '-d', fname_dest, '-w', list_fname_warp[i_file], '-x', param.interp, '-o', 'src_' + str(i_file) + '_template_partialVolume.nii.gz' ]) # open data data[:, :, :, i_file] = Image('src_' + str(i_file) + '_template.nii.gz').data partial_volume[:, :, :, i_file] = Image('src_' + str(i_file) + '_template_partialVolume.nii.gz').data i_file += 1 # merge files using partial volume information (and convert nan resulting from division by zero to zeros) data_merge = np.divide(np.sum(data * partial_volume, axis=3), np.sum(partial_volume, axis=3)) data_merge = np.nan_to_num(data_merge) # write result in file nii_dest.data = data_merge nii_dest.save(param.fname_out) # remove temporary folder if param.rm_tmp: rmtree(path_tmp)
def main(argv=None): """ Main function :param argv: :return: """ parser = get_parser() arguments = parser.parse_args(argv) verbose = arguments.v set_global_loglevel(verbose=verbose) dim_list = ['x', 'y', 'z', 't'] fname_in = arguments.i fname_out = arguments.o output_type = arguments.type # Open file(s) im = Image(fname_in) data = im.data # 3d or 4d numpy array dim = im.dim # run command if arguments.otsu is not None: param = arguments.otsu data_out = sct_math.otsu(data, param) elif arguments.adap is not None: param = arguments.adap data_out = sct_math.adap(data, param[0], param[1]) elif arguments.otsu_median is not None: param = arguments.otsu_median data_out = sct_math.otsu_median(data, param[0], param[1]) elif arguments.thr is not None: param = arguments.thr data_out = sct_math.threshold(data, param) elif arguments.percent is not None: param = arguments.percent data_out = sct_math.perc(data, param) elif arguments.bin is not None: bin_thr = arguments.bin data_out = sct_math.binarize(data, bin_thr=bin_thr) elif arguments.add is not None: data2 = get_data_or_scalar(arguments.add, data) data_concat = sct_math.concatenate_along_4th_dimension(data, data2) data_out = np.sum(data_concat, axis=3) elif arguments.sub is not None: data2 = get_data_or_scalar(arguments.sub, data) data_out = data - data2 elif arguments.laplacian is not None: sigmas = arguments.laplacian if len(sigmas) == 1: sigmas = [sigmas for i in range(len(data.shape))] elif len(sigmas) != len(data.shape): printv( parser.error( 'ERROR: -laplacian need the same number of inputs as the number of image dimension OR only one input' )) # adjust sigma based on voxel size sigmas = [sigmas[i] / dim[i + 4] for i in range(3)] # smooth data data_out = sct_math.laplacian(data, sigmas) elif arguments.mul is not None: data2 = get_data_or_scalar(arguments.mul, data) data_concat = sct_math.concatenate_along_4th_dimension(data, data2) data_out = np.prod(data_concat, axis=3) elif arguments.div is not None: data2 = get_data_or_scalar(arguments.div, data) data_out = np.divide(data, data2) elif arguments.mean is not None: dim = dim_list.index(arguments.mean) if dim + 1 > len( np.shape(data)): # in case input volume is 3d and dim=t data = data[..., np.newaxis] data_out = np.mean(data, dim) elif arguments.rms is not None: dim = dim_list.index(arguments.rms) if dim + 1 > len( np.shape(data)): # in case input volume is 3d and dim=t data = data[..., np.newaxis] data_out = np.sqrt(np.mean(np.square(data.astype(float)), dim)) elif arguments.std is not None: dim = dim_list.index(arguments.std) if dim + 1 > len( np.shape(data)): # in case input volume is 3d and dim=t data = data[..., np.newaxis] data_out = np.std(data, dim, ddof=1) elif arguments.smooth is not None: sigmas = arguments.smooth if len(sigmas) == 1: sigmas = [sigmas[0] for i in range(len(data.shape))] elif len(sigmas) != len(data.shape): printv( parser.error( 'ERROR: -smooth need the same number of inputs as the number of image dimension OR only one input' )) # adjust sigma based on voxel size sigmas = [sigmas[i] / dim[i + 4] for i in range(3)] # smooth data data_out = sct_math.smooth(data, sigmas) elif arguments.dilate is not None: if arguments.shape in ['disk', 'square'] and arguments.dim is None: printv( parser.error( 'ERROR: -dim is required for -dilate with 2D morphological kernel' )) data_out = sct_math.dilate(data, size=arguments.dilate, shape=arguments.shape, dim=arguments.dim) elif arguments.erode is not None: if arguments.shape in ['disk', 'square'] and arguments.dim is None: printv( parser.error( 'ERROR: -dim is required for -erode with 2D morphological kernel' )) data_out = sct_math.erode(data, size=arguments.erode, shape=arguments.shape, dim=arguments.dim) elif arguments.denoise is not None: # parse denoising arguments p, b = 1, 5 # default arguments list_denoise = (arguments.denoise).split(",") for i in list_denoise: if 'p' in i: p = int(i.split('=')[1]) if 'b' in i: b = int(i.split('=')[1]) data_out = sct_math.denoise_nlmeans(data, patch_radius=p, block_radius=b) elif arguments.symmetrize is not None: data_out = (data + data[list(range(data.shape[0] - 1, -1, -1)), :, :]) / float(2) elif arguments.mi is not None: # input 1 = from flag -i --> im # input 2 = from flag -mi im_2 = Image(arguments.mi) compute_similarity(im, im_2, fname_out, metric='mi', metric_full='Mutual information', verbose=verbose) data_out = None elif arguments.minorm is not None: im_2 = Image(arguments.minorm) compute_similarity(im, im_2, fname_out, metric='minorm', metric_full='Normalized Mutual information', verbose=verbose) data_out = None elif arguments.corr is not None: # input 1 = from flag -i --> im # input 2 = from flag -mi im_2 = Image(arguments.corr) compute_similarity(im, im_2, fname_out, metric='corr', metric_full='Pearson correlation coefficient', verbose=verbose) data_out = None # if no flag is set else: data_out = None printv( parser.error( 'ERROR: you need to specify an operation to do on the input image' )) if data_out is not None: # Write output nii_out = Image(fname_in) # use header of input file nii_out.data = data_out nii_out.save(fname_out, dtype=output_type) # TODO: case of multiple outputs # assert len(data_out) == n_out # if n_in == n_out: # for im_in, d_out, fn_out in zip(nii, data_out, fname_out): # im_in.data = d_out # im_in.absolutepath = fn_out # if arguments.w is not None: # im_in.hdr.set_intent('vector', (), '') # im_in.save() # elif n_out == 1: # nii[0].data = data_out[0] # nii[0].absolutepath = fname_out[0] # if arguments.w is not None: # nii[0].hdr.set_intent('vector', (), '') # nii[0].save() # elif n_out > n_in: # for dat_out, name_out in zip(data_out, fname_out): # im_out = nii[0].copy() # im_out.data = dat_out # im_out.absolutepath = name_out # if arguments.w is not None: # im_out.hdr.set_intent('vector', (), '') # im_out.save() # else: # printv(parser.usage.generate(error='ERROR: not the correct numbers of inputs and outputs')) # display message if data_out is not None: display_viewer_syntax([fname_out], verbose=verbose) else: printv('\nDone! File created: ' + fname_out, verbose, 'info')
def main(argv=None): parser = get_parser() arguments = parser.parse_args(argv) verbose = arguments.v set_loglevel(verbose=verbose) fname_input1 = arguments.i fname_input2 = arguments.d tmp_dir = tmp_create() # create tmp directory tmp_dir = os.path.abspath(tmp_dir) # copy input files to tmp directory fname_input1_tmp = 'tmp1_' + ''.join(extract_fname(fname_input1)[1:]) fname_input2_tmp = 'tmp2_' + ''.join(extract_fname(fname_input2)[1:]) copy(fname_input1, os.path.join(tmp_dir, fname_input1_tmp)) copy(fname_input2, os.path.join(tmp_dir, fname_input2_tmp)) fname_input1 = fname_input1_tmp fname_input2 = fname_input2_tmp curdir = os.getcwd() os.chdir(tmp_dir) # go to tmp directory im_1 = Image(fname_input1) im_2 = Image(fname_input2) if arguments.bin is not None: im_1.data = binarize(im_1.data, 0) fname_input1_bin = add_suffix(fname_input1, '_bin') im_1.save(fname_input1_bin, mutable=True) im_2.data = binarize(im_2.data, 0) fname_input2_bin = add_suffix(fname_input2, '_bin') im_2.save(fname_input2_bin, mutable=True) # Use binarized images in subsequent steps fname_input1 = fname_input1_bin fname_input2 = fname_input2_bin # copy header of im_1 to im_2 im_2.header = im_1.header im_2.save() cmd = ['isct_dice_coefficient', fname_input1, fname_input2] if vars(arguments)["2d_slices"] is not None: cmd += ['-2d-slices', str(vars(arguments)["2d_slices"])] if arguments.b is not None: bounding_box = arguments.b cmd += ['-b'] + bounding_box if arguments.bmax is not None and arguments.bmax == 1: cmd += ['-bmax'] if arguments.bzmax is not None and arguments.bzmax == 1: cmd += ['-bzmax'] if arguments.o is not None: path_output, fname_output, ext = extract_fname(arguments.o) cmd += ['-o', fname_output + ext] rm_tmp = bool(arguments.r) # # Computation of Dice coefficient using Python implementation. # # commented for now as it does not cover all the feature of isct_dice_coefficient # #from spinalcordtoolbox.image import Image, compute_dice # #dice = compute_dice(Image(fname_input1), Image(fname_input2), mode='3d', zboundaries=False) # #printv('Dice (python-based) = ' + str(dice), verbose) status, output = run_proc(cmd, verbose, is_sct_binary=True) os.chdir(curdir) # go back to original directory # copy output file into original directory if arguments.o is not None: copy(os.path.join(tmp_dir, fname_output + ext), os.path.join(path_output, fname_output + ext)) # remove tmp_dir if rm_tmp: rmtree(tmp_dir) printv(output, verbose)