import os
import nibabel as nib
import numpy as np
from overlay_ims import overlay_image, save_overlay_im
from PIL import Image
def make_montage(sname_ims, sname_mont='montage.png', crop=None):
"""
Make images
Args:
sname_ims (list): list of image names
sname_mont (str): name of output file
crop (None or list of length 4): Crop coordinates (x, y, x_width, y_width)
Returns:
"""
rows = len(sname_ims)
y = 0
for i in range(rows):
x = 0
# Load first row of images
im = Image.open(sname_ims[i])
msk = Image.open(sname_msks[i])
seg = Image.open(sname_segs[i])
diff = Image.open(sname_diffs[i])
# Apply crops
if crop:
bbox = (crop[0], crop[1], im.width + crop[2], im.height + crop[3])
im = im.crop(bbox)
msk = msk.crop(bbox)
seg = seg.crop(bbox)
diff = diff.crop(bbox)
print(im.height)
if i == 0:
mont = Image.new('RGB', size=(im.width * 4, im.height * rows))
mont.paste(im, (x, y))
x += im.width
mont.paste(msk, (x, y))
x += im.width
mont.paste(seg, (x, y))
x += im.width
mont.paste(diff, (x, y))
# Increment row
y += im.height
# Save montage image
sname_mont = os.path.join(save_path, sname_mont)
mont.save(sname_mont)
def make_montage_init(sname_ims, sname_mont='montage.png'):
"""
Make a montage with one row
Args:
sname_ims (str): montage save name
sname_mont (list): lits of image files
Returns:
"""
# Make montage
cols = len(sname_ims)
x = 0
y = 0
for i in range(cols):
# Load first row of images
im = Image.open(sname_ims[i])
if i == 0:
mont = Image.new('RGB', size=(im.width * cols, im.height))
mont.paste(im, (x, y))
x += im.width
# Save montage image
sname_mont = os.path.join(save_path, sname_mont)
mont.save(sname_mont)
def load_label_show(files, sz):
"""
Load and prep label file
Args:
files (str): label filename
sz (list of length 2): size of MR image (may not always match label)
Returns:
"""
from skimage.transform import resize
from scipy.ndimage.morphology import binary_closing
# Load image
label = nib.load(files).get_data().astype('float32').squeeze()
# Resize image
label /= label.max()
label = resize(label, (sz[0], sz[1]))
# Make binary
label[label < 0.5] = 0
label[label > 0.5] = 1
# Close the mask
# label = binary_closing(label, structure=np.ones(shape=(3, 3, 3)))
return 255 * label.astype(np.uint8)
"""
Goal: remake segmentation images to show a control and a PD1 animal before and after segmentation.
"""
# Set up save paths
save_path = '/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/Analysis/Segmentation_images'
snames = [
['control_1',
'control_2'],
['pd1_1',
'pd1_2']
]
# Data paths
image_files = [
['/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520552/20180510/T2_cor.nii.gz',
'/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520552/20180517/T2_cor.nii.gz'], # Control
['/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180503/T2_cor.nii.gz',
'/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180510/T2_cor.nii.gz']
]
seg_files = [
['/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520552/20180510/tumor_seg.nii.gz',
'/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520552/20180517/tumor_seg.nii.gz'],
['/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180503/tumor_seg.nii.gz',
'/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180510/tumor_seg.nii.gz']
]
mask_files = [
['/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/520552/Seg_520552-1-label.nii',
'/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/520552/Seg_520552-2-label.nii'],
['/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/520457/Seg_520457-1-label.nii',
'/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/520457/Seg_520457-2-label.nii']
]
slices = [[27, 19],
[27, 19]
]
# Change paths if in windows
if 'nt' in os.name:
save_path = save_path.replace('/media/matt/Seagate Expansion Drive/', 'E://')
for i in range(len(image_files)):
image_files[i] = [i.replace('/media/matt/Seagate Expansion Drive/', 'E://') for i in image_files[i]]
seg_files[i] = [i.replace('/media/matt/Seagate Expansion Drive/', 'E://') for i in seg_files[i]]
mask_files[i] = [i.replace('/media/matt/Seagate Expansion Drive/', 'E://') for i in mask_files[i]]
if not os.path.exists(save_path):
os.mkdir(save_path)
# Colormap
colormap_mask = ['black', 'red']
colormap_seg = ['black', 'green']
colormap_comb = ['black', 'red', 'green']
# Histogram clipping
top_pct = 0.05 / 100
bot_pct = 15 / 100
sname_ims, sname_msks, sname_segs, sname_diffs = [], [], [], []
for i in range(len(image_files)):
for ii in range(len(image_files[i])):
# Read image
im = nib.load(image_files[i][ii]).get_data().astype(np.float)
msk = nib.load(mask_files[i][ii]).get_data().astype(np.float)
seg = nib.load(seg_files[i][ii]).get_data().astype(np.float)
# Scale the image
im_s = sorted(im.reshape(-1))
top_thresh = im_s[-round(len(im_s) * top_pct)]
bot_thresh = im_s[round(len(im_s) * bot_pct)]
im[im > top_thresh] = top_thresh
im[im < bot_thresh] = bot_thresh
# Select slice
s = slices[i][ii]
im = im[:, :, s].T
msk = msk[:, :, s].T
seg = seg[:, :, s].T
# Normalize
# Set up save name
save_name = os.path.join(save_path, snames[i][ii])
# Create overlay image - no masks
sname_im = save_name + 'im.png'
over_im = overlay_image(base_image=im,
overlay_image=np.zeros_like(msk),
colormap=colormap_mask,
sname=sname_im
)
# Create overlay image - mask
sname_msk = save_name + 'msk.png'
over_im = overlay_image(base_image=im,
overlay_image=msk,
colormap=colormap_mask,
sname=sname_msk
)
# Create overlay image - segmentation
sname_seg = save_name + 'seg.png'
over_im = overlay_image(base_image=im,
overlay_image=seg,
colormap=colormap_seg,
sname=sname_seg
)
# Create overlay image - combined
# Green = false positive, 1.0
# Red = false negative, 0.5
# False positives
fp = seg - msk
fp[fp < 0] = 0
# False negatives
fn = msk - seg
fn[fn < 0] = 0
# Combined
comb = fp + 0.5 * fn
sname_diff = save_name + 'diff.png'
over_im = overlay_image(base_image=im,
overlay_image=comb,
colormap=colormap_comb,
sname=sname_diff
)
sname_ims.append(sname_im)
sname_msks.append(sname_msk)
sname_segs.append(sname_seg)
sname_diffs.append(sname_diff)
make_montage(sname_ims, crop=(35, 10, -35, -70))
""" Initial images """
# Show modal images and segmentation
files = ['/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/3/180420-3b-2_T1.nii',
'/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/3/180420-3b-2_T1wC.nii',
'/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/3/180420-3B-2_T2.nii']
ims = []
seg_file = '/media/matt/Seagate Expansion Drive/MR Data/MR_Images_Sarcoma/3/180420-3B-2-label.nii'
snames = ['T1', 'T1C', 'T2', 'seg']
snames = [os.path.join(save_path, i + '_2.png') for i in snames]
slice = 33
# Change paths if in windows
if 'nt' in os.name:
seg_file = seg_file.replace('/media/matt/Seagate Expansion Drive/', 'E://')
files = [i.replace('/media/matt/Seagate Expansion Drive/', 'E://') for i in files]
for file, sname in zip(files, snames[:-1]):
# Load the file and select slice
im = nib.load(file).get_data().astype(np.float)
im = im[:, :, slice].T
# Remove outliers
im_s = sorted(im.reshape(-1))
top_thresh = im_s[-round(len(im_s) * top_pct)]
bot_thresh = im_s[round(len(im_s) * bot_pct)]
im[im > top_thresh] = top_thresh
im[im < bot_thresh] = bot_thresh
# Convert to 8 bit
im -= im.min()
im /= im.max()
im *= 255
im = im.astype(np.uint8)
# Save image
save_overlay_im(im, sname)
# Save the segmenation image
seg = nib.load(seg_file).get_data().astype(np.uint8)
seg = seg[:, :, slice].T
seg -= seg.min()
seg = seg / seg.max()
seg *= 255
seg = seg.astype(np.uint8)
# Apply colormap
seg = np.repeat(seg[:, :, np.newaxis], 3, 2)
seg[:, :, 1:] = 0
# Save
save_overlay_im(seg, snames[-1])
# Save montage
make_montage_init(snames, sname_mont='mont_init1.png')
# Save second sample images
files = ['/media/matt/Seagate Expansion Drive/MR Data/SarcomaSegmentations/Mouse IV/Mouse IV T1 FLASH.nii',
'/media/matt/Seagate Expansion Drive/MR Data/SarcomaSegmentations/Mouse IV/Mouse IV T1FLASH with Contrast.nii',
'/media/matt/Seagate Expansion Drive/MR Data/SarcomaSegmentations/Mouse IV/Bias Corrected Images/Mouse IV T2TurboRARE New Bias Correction.nii']
seg_file = '/media/matt/Seagate Expansion Drive/MR Data/SarcomaSegmentations/Mouse IV/Mouse IV ROI Black and White Volume.nii'
snames = ['T1', 'T1C', 'T2', 'seg']
snames = [os.path.join(save_path, i + '.png') for i in snames]
slice = 25
# Change paths if in windows
if 'nt' in os.name:
seg_file = seg_file.replace('/media/matt/Seagate Expansion Drive/', 'E://')
files = [i.replace('/media/matt/Seagate Expansion Drive/', 'E://') for i in files]
for file, sname in zip(files, snames[:-1]):
# Load the file and select slice
im = nib.load(file).get_data().astype(np.float)
sz = im.shape
im = im[:, :, slice].T
# Remove outliers
im_s = sorted(im.reshape(-1))
top_thresh = im_s[-round(len(im_s) * top_pct)]
bot_thresh = im_s[round(len(im_s) * bot_pct)]
im[im > top_thresh] = top_thresh
im[im < bot_thresh] = bot_thresh
# Convert to 8 bit
im -= im.min()
im /= im.max()
im *= 255
im = im.astype(np.uint8)
# Save image
save_overlay_im(im, sname)
# Save the segmenation image
seg = load_label_show(seg_file, sz)
seg = seg[:, :, slice].T
# Apply colormap
seg = np.repeat(seg[:, :, np.newaxis], 3, 2)
seg[:, :, 1:] = 0
# Save
save_overlay_im(seg, snames[-1])
# Save montage
make_montage_init(snames, sname_mont='mont_init2.png')
""" Show tumor, edge, and bed segmentations """
from Radiomics.radiomic_maps_190501 import dilate_masks
mask_tumor = '/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180510/tumor_seg.nii.gz'
image_name = '/media/matt/Seagate Expansion Drive/b7TData_19/b7TData/Results/K520457/20180510/T2_cor.nii.gz'
snames = ['area_tumor.png', 'area_edge.png', 'area_bed.png']
snames = [os.path.join(save_path, i) for i in snames]
descriptor = 'area'
slices = [29]
top_pct = 1 / 100
bot_pct = 15 / 100
# Compute masks
mask_bed_file = dilate_masks(mask_tumor, image_name, save_path, diff=False, name=descriptor)
mask_edge_file = dilate_masks(mask_tumor, image_name, save_path, diff=True, name=descriptor)
# Load image data
im = nib.load(image_name).get_data().astype(np.float32)
mask_tumor = nib.load(mask_tumor).get_data().astype(np.uint8)
mask_bed = nib.load(mask_bed_file).get_data().astype(np.uint8)
mask_edge = nib.load(mask_edge_file).get_data().astype(np.uint8)
# Delete dilated masks
# os.remove(mask_bed_file)
# os.remove(mask_edge_file)
# Preprocess image
# Remove outliers
# Remove by pct
im_s = sorted(im.reshape(-1))
top_thresh = im_s[-round(len(im_s) * top_pct)]
bot_thresh = im_s[round(len(im_s) * bot_pct)]
im[im > top_thresh] = top_thresh
im[im < bot_thresh] = bot_thresh
# Convert to 8 bit
# im -= im.min()
# im /= im.max()
# im *= 255
# im = im.astype(np.uint8)
colormap_seg2 = ['black', 'green']
rows = range(20, im.shape[0] - 20)
cols = range(10, im.shape[1] - 10)
# Make image overlays
for i, mask in enumerate([mask_tumor, mask_edge, mask_bed]):
overlay_image(base_image=im[rows, :, slices[0]].T,
overlay_image=mask[rows, :, slices[0]].T,
colormap=colormap_seg2,
sname=snames[i])
# Combine images into a montage
make_montage_init(snames, sname_mont='area_montage.png')
