def divide_certain(slide_path: str, out_dir: str, logger, width=96) -> None:
"""The origin slide is too large, the function can segment the large one into small tiles.
In this project, we set the height equals to the width. It aims to tile images in 10X power using
less resource as possibel"""
slide, level, base10, dimension_ref, d_step, size = divide_prepare(
slide_path, width, logger)
# begin segment tiles
cwp = os.getcwd()
case_name = get_name(slide_path)
out_path = os.path.join(cwp, out_dir, case_name)
os.makedirs(out_path, exist_ok=True)
# set start points of tiles
height = width
x_bound = dimension_ref[0] - width
y_bound = dimension_ref[1] - height
for x in tqdm(range(0, x_bound, d_step)):
for y in range(0, y_bound, d_step):
loc = (x, y)
# print(loc, level, size)
small_image = slide.read_region(location=loc,
level=level,
size=size)
if is_useless(small_image):
continue
if not base10:
small_image = small_image.resize((width, height))
fp = os.path.join(out_path,
'{}-{}-{}-{}.tiff'.format(x, y, width, height))
small_image.save(fp)
def __init__(self,filename,dirname):
self.filename=tools.get_name(filename)
self.variables={}
self.dirname=dirname
self.comments={}
self.order_variables=[]
def __init__(self, filename, dirname):
self.filename = tools.get_name(filename)
self.variables = {}
self.dirname = dirname
self.comments = {}
self.order_variables = []
def combined_graph(scores, episode_numbers, name, coordinates=None, linears=None, scatter=False):
"""
method prints point graph and
interpolation graph with gaussian filter of learning progress
"""
if linears is not None:
for key, value in linears.items():
plt.plot([0, episode_numbers[-1]], [key, value], 'k-', linewidth=0.8)
if scatter:
plt.plot(episode_numbers, scores, 'ro', color='goldenrod', markersize=1)
score_gf = gaussian_filter(scores, sigma=0.01791*episode_numbers[-1])
plt.plot(episode_numbers, score_gf, color='teal', linewidth=1)
plt.ylabel("Score")
plt.xlabel("Episode")
plt.xlim([0,coordinates[0]])
if min(scores) < 0:
plt.ylim([min(scores),coordinates[1]])
else:
plt.ylim([0,coordinates[1]])
name = get_name(name)
plt.savefig("./{}" .format(name), bbox_inches='tight')
plt.clf()
print("[Graph of learning progress visualization was saved to \"./{}\".]" .format(name))
def test_get_name(self):
"""
test for get_name
"""
expected = "4-Hydroxybenzaldehyde"
Cnumber = "C00002657"
actual = tools.get_name(Cnumber)
time.sleep(2)
self.assertEqual(actual, expected)
def visualize_model(model, plot_mdl=[True, False]):
"""
method prints model to stdout and pdf
"""
if plot_mdl[0]:
model.summary()
if plot_mdl[1]:
name = get_name("model")
plot_model(model,
to_file=name,
show_shapes=True,
show_layer_names=False)
def batch_tiling(path, out_dir, logger):
filter_func = None
# filter_func = lambda x:get_name(x) not in cache
slides = list(filter(filter_func, get_files(path)))
for slide_path in slides:
name = get_name(slide_path)
try:
logger.info(f'start {name}')
divide(slide_path, out_dir)
except:
logger.exception(f'{name} encountered error in batch')
def eval_using_area_under_curve(generation_directory):
# get the name and the area under curve for each child in this directory
areas = []
for dir in [x for x in generation_directory.iterdir() if x.is_dir()]:
area_file = dir / 'auc.txt'
area_val = float(area_file.read_text())
child_name = get_name(dir)
dna_file = dir / 'dna.dna'
dna = dna_file.read_text()
areas.append( (area_val , dna , child_name))
areas.sort(key=lambda x: x[0])
generate_generation_report(generation_directory , areas)
return areas[0][1] , areas[1][1]
def divide(slide_path: str,
out_dir: str,
level=0,
width_rel=96,
mag=10) -> None:
"""The origin slide is too large, the function can segment the large one into small tiles.
In this project, we set the height equals to the width.
Slide_path: the path of the target slide;
level: varying definition of images, 0 is the largest, int;
width_rel: the width and the length of output images, int.
mag: the magnitude or the object power, float"""
# Read slide. level 0 is mag X40 or X20.
# here we downsize ((level 0 's mag / mag)**2) times
large_image = openslide.OpenSlide(slide_path)
ori_mag = int(large_image.properties.get('openslide.objective-power'))
time = ori_mag / mag
tile = int(width_rel * time)
# get reference and target location, use a reference level instead of the maxiumn power level may make reduce the cose of resize
dimensions = large_image.level_dimensions
dimension_ref = dimensions[0]
dimension = dimensions[level]
ratio = dimension[0] / dimension_ref[0]
# begin segment tiles
case_name = get_name(slide_path)
# print(case_name)
out_path = os.path.join(out_dir, case_name)
os.makedirs(out_path, exist_ok=True)
# calculate individual size
height_rel = width_rel
width = int(tile * ratio)
height = int(width)
size = (width, height)
x_bound = dimension_ref[0] - width
y_bound = dimension_ref[1] - height
for x in tqdm(range(0, x_bound, tile)):
for y in (range(0, y_bound, tile)):
# locate start point
loc = (x, y)
# get the small image
small_image = large_image.read_region(location=loc,
level=level,
size=size)
# filter the useless image
if is_useless(small_image):
continue
# save the small image
resize_image = small_image.resize((width_rel, height_rel))
fp = os.path.join(out_path,
'{}-{}-{}-{}.tiff'.format(x, y, width, height))
resize_image.save(fp)
def heat_map(array, graph_name, axes):
"""
method creates heatmap
"""
fig, ax = plt.subplots()
im = ax.imshow(array, cmap=cm.YlOrRd)
if not axes:
im.axes.get_yaxis().set_visible(False)
im.axes.get_xaxis().set_visible(False)
fig.tight_layout()
graph_name = get_name(graph_name)
plt.savefig("./{}".format(graph_name), bbox_inches='tight')
print("[Heatmap was made.]")
def visualize_model(self, model):
"""
method prints model to stdout and pdf
"""
if not self.plotted:
if self.plot_mdl[0]:
model.summary()
if self.plot_mdl[1]:
name = get_name("model")
plot_model(model,
to_file=name,
show_shapes=True,
show_layer_names=False)
print("[PDF blueprint of neural network was saved.]")
self.plotted = True
def import_settings(self, filename):
comment = ""
f = open(filename)
for line in f:
line = line.strip()
if (line[0] == "#"):
comment = comment + line
continue
line = line.split('=')
self.variables[line[0]] = line[1]
self.order_variables.append(line[0])
self.comments[line[0]] = comment
comment = ""
filename = tools.get_name(filename)
self.filename = filename
f.close()
def import_settings(self,filename):
comment=""
f=open(filename)
for line in f :
line=line.strip()
if (line[0] == "#") :
comment=comment+line
continue
line=line.split('=')
self.variables[line[0]]=line[1]
self.order_variables.append(line[0])
self.comments[line[0]]=comment
comment=""
filename=tools.get_name(filename)
self.filename=filename
f.close()
def set_name(self, flat):
name = get_name(flat)
self.name_line.setText(str(name))
def convert(img_p: str, out_path: str) -> None:
img = cv2.imread(img_p)
name = get_name(img_p)
cv2.imwrite(os.path.join(out_path, f'{name}.tiff'), img)
def set_name(self, flat):
name = get_name(flat)
self.name_value.setText(str(name))
if self.save_button is not None:
self.save_button.setEnabled(True)