def build_annotated(self):
thumb = create_section_image(self.section, self.experiment_id,
self.directory, self.full_data,
self.bboxes)
_, r, _ = detect_brain(cv2.cvtColor(thumb, cv2.COLOR_BGR2GRAY))
cv2.imwrite(self.annotated_url, thumb[r.y:r.y + r.h,
r.x:r.x + r.w])
def calculate_global_parameters(self, density3d_maps, dense_masks, cells):
relevant_sections = cells.section.unique()
relevant_sections.sort()
globs_per_section = infinite_dict()
for section in relevant_sections:
self.logger.debug(f"Calculating globals for section {section} of {self.id}...")
section_seg_data = self.seg_data[:, :, section]
for region, (start, end, shape, (mask_y, mask_x, mask_section)) in dense_masks.items():
if start <= section <= end:
if shape[0] < section_seg_data.shape[0]:
ratio = section_seg_data.shape[0] // shape[0]
deltas_x = np.array(([i for i in range(ratio)] * ratio) * len(mask_x))
deltas_y = np.array([[i] * ratio for i in range(ratio)] * len(mask_x))
mask_y = np.kron(mask_y * ratio, np.ones((1, ratio * ratio))).flatten() + deltas_y
mask_x = np.kron(mask_x * ratio, np.ones((1, ratio * ratio))).flatten() + deltas_x
elif shape[0] > section_seg_data.shape[0]:
ratio = shape[0] // section_seg_data.shape[0]
section_seg_data = np.kron(section_seg_data,
np.ones((ratio, ratio), dtype=section_seg_data.dtype))
relevant_cells = mask_section == (section - start)
section_seg_data[mask_y[relevant_cells], mask_x[relevant_cells]] = region
section_density_map = density3d_maps[2][:, :, section - density3d_maps[0]]
if section_density_map.shape[0] > section_seg_data.shape[0]:
ratio = section_density_map.shape[0] // section_seg_data.shape[0]
section_seg_data = np.kron(section_seg_data, np.ones((ratio, ratio), dtype=section_seg_data.dtype))
elif section_density_map.shape[0] < section_seg_data.shape[0]:
ratio = section_seg_data.shape[0] // section_density_map.shape[0]
section_density_map = np.kron(section_density_map,
np.ones((ratio, ratio), dtype=section_density_map.dtype))
_, bbox, _ = detect_brain((section_seg_data != 0).astype(np.uint8) * 255)
center_x = bbox.x + bbox.w // 2
scale_factor = (0.35 * 64) / (section_seg_data.shape[0] / self.seg_data.shape[0])
relevant_regions = np.intersect1d(np.unique(section_seg_data),
cells[cells.section == section].structure_id.unique())
for region in relevant_regions:
region_cells = np.where(section_seg_data == region)
globs_per_section[region][section]['region_area'] = region_cells[0].shape[0] * (scale_factor ** 2)
globs_per_section[region][section]['region_area_left'] = np.where(region_cells[1] < center_x)[0].shape[
0] * (
scale_factor ** 2)
globs_per_section[region][section]['region_area_right'] = \
np.where(region_cells[1] >= center_x)[0].shape[
0] * (
scale_factor ** 2)
densities = section_density_map[region_cells].flatten()
densities = (densities[densities != 0], len(densities))
globs_per_section[region][section]['density3d'] = densities
return globs_per_section
def get_brain_bbox_and_image(bboxes, directory, experiment_id, section, image_needed, scale=4):
thumb = cv2.imread(f"{directory}/thumbnail-{experiment_id}-{section}.jpg", cv2.IMREAD_GRAYSCALE)
_, brain_bbox, _ = detect_brain(thumb)
thumb = cv2.resize(thumb, (0, 0), fx=64 // scale, fy=64 // scale)
if image_needed:
for bbox in bboxes[section]:
x, y, w, h = bbox.scale(64)
image = cv2.imread(f'{directory}/full-{experiment_id}-{section}-{x}_{y}_{w}_{h}.jpg',
cv2.IMREAD_GRAYSCALE)
x, y, w, h = bbox.scale(64 // scale)
thumb[y: y + h, x: x + w] = cv2.resize(image, (0, 0), fx=1.0 / scale, fy=1.0 / scale)
return thumb, brain_bbox.pad(5, 5).scale(64)
def __init__(self, experiment_id, section, data, base_time, input_dir,
structure_tree):
self.structure_tree = structure_tree
self.experiment_id = experiment_id
self.data = data
# self.hist_button = widgets.Button(description=f"Show detailed histograms")
# self.hist_button.layout.width = 'auto'
# self.hist_button.on_click(self.clicked)
self.output = widgets.Output()
self.section = section
display(Markdown('---'))
if self.section != 'totals':
self.section = int(section)
display(
widgets.Label(
f"Experiment {self.experiment_id}, section {self.section}")
)
self.annotated_button_bar = self.AnnotationsButtonBar(
self.experiment_id, self.data, self.section, base_time,
input_dir, self.output)
display(self.output)
display(widgets.HBox((self.annotated_button_bar, )))
if os.path.isfile(
f'{input_dir}/{experiment_id}/thumbnail-{self.experiment_id}-{section}.jpg'
):
thumb = Image.open(
f'{input_dir}/{experiment_id}/thumbnail-{self.experiment_id}-{section}.jpg'
)
_, rect, _ = detect_brain(
np.array(thumb.convert('LA'))[:, :, 0])
thumb = thumb.crop((
*(rect.corners()[0]),
*(rect.corners()[1]),
))
else:
thumb = None
else:
display(widgets.Label(f"Experiment {self.experiment_id}, totals"))
display(self.output)
display(
widgets.HBox(
(self.HeatmapAndPatchButtons(experiment_id, base_time,
input_dir), )))
thumb = None
with self.output:
# plot_section_violin_diagram(self.data, structure_tree, thumb)
plt.imshow(thumb)
plt.axis('off')
plt.show()
def get_brain_metrics(self, section):
thumbnail_file_name = os.path.join(self.directory, f'thumbnail-{self.id}-{section}.jpg')
thumbnail = cv2.imread(thumbnail_file_name, cv2.IMREAD_GRAYSCALE)
brain_mask, bbox, ctrs = detect_brain(thumbnail)
brain_area = sum([Polygon(ctr.squeeze()).area for ctr in ctrs])
return brain_area, bbox