def prepare_patient_images(patient_id, intermediate_crop=0):
file_lst = []
prefix = str(patient_id).rjust(4, '0')
src_files = helpers.get_files(settings.BASE_PREPROCESSEDIMAGES_DIR, prefix + "*.png")
patient_dir = helpers.get_pred_patient_dir(patient_id)
helpers.create_dir_if_not_exists(patient_dir)
patient_img_dir = helpers.get_pred_patient_img_dir(patient_id)
helpers.create_dir_if_not_exists(patient_img_dir)
helpers.delete_files(patient_img_dir, "*.png")
dummy = numpy.zeros((settings.TARGET_SIZE, settings.TARGET_SIZE))
cv2.imwrite(patient_img_dir + "dummy_overlay.png", dummy)
for src_path in src_files:
file_name = ntpath.basename(src_path)
org_img = cv2.imread(src_path, cv2.IMREAD_GRAYSCALE)
cropped_img = helpers.prepare_cropped_sax_image(org_img, clahe=True, intermediate_crop=intermediate_crop, rotate=0)
if SCALE_SIZE is not None:
cropped_img = cv2.resize(cropped_img, (SCALE_SIZE, SCALE_SIZE), interpolation=cv2.INTER_AREA)
cv2.imwrite(patient_img_dir + file_name, cropped_img)
file_lst.append([file_name, "dummy_overlay.png"])
with open(patient_img_dir + "pred.lst", "wb") as f:
writer = csv.writer(f, delimiter='\t')
writer.writerows(file_lst)
def compute_volumes(patient_id, model_name, debug_info=False):
patient_dir = helpers.get_pred_patient_dir(patient_id)
min_areas = pandas.read_csv(patient_dir + "\\areas_" + model_name + ".csv", sep=";")
columns = list(min_areas)
# diastole_col = ""
diastole_pixels = 0
systole_pixels = 999999
diastole_max = 0
systole_max = 999999
for column in columns:
if not column.startswith("fr"):
continue
# if False:
# pixel_max = min_areas[column].max()
# if pixel_max > diastole_max:
# diastole_max = min_areas[column].max()
# diastole_col = column
#
# pixel_max = min_areas[column].max()
# if pixel_max < systole_max:
# systole_max = pixel_max
# systole_col = column
# else:
value_list = min_areas[column].values.tolist()
value_list.sort(reverse=True)
pixel_sum = sum(value_list[:200])
#pixel_sum = min_areas[column].sum()
if pixel_sum > diastole_pixels:
diastole_pixels = pixel_sum
diastole_col = column
if pixel_sum < systole_pixels:
systole_pixels = pixel_sum
systole_col = column
if debug_info:
current_debug_line.append(str(diastole_col))
current_debug_line.append(str(systole_col))
# min_areas["diastole_vol"] = min_areas[diastole_col] * min_areas["slice_dist"]
# min_areas["systole_vol"] = min_areas[systole_col] * min_areas["slice_dist"]
dist_col = "slice_dist"
min_areas_selection = min_areas[["slice", "slice_thickness", "slice_location", "time", dist_col]].copy() # , "diastole_vol", "systole_vol"
min_areas_selection["diastole"] = min_areas[diastole_col].values
min_areas_selection["diastole_vol"] = (min_areas_selection["diastole"] * min_areas_selection[dist_col]).values
min_areas_selection["diastole_conf"] = min_areas[diastole_col.replace("fr", "co")].values
min_areas_selection["systole"] = min_areas[systole_col].values
min_areas_selection["systole_vol"] = (min_areas_selection["systole"] * min_areas_selection[dist_col]).values
min_areas_selection["systole_conf"] = min_areas[systole_col.replace("fr", "co")].values
min_areas_selection.to_csv(patient_dir + "\\areas_dia_sys_" + model_name + ".csv", sep=";")
dia_frame = diastole_col.replace("fr_", "")
sys_frame = systole_col.replace("fr_", "")
diastole_vol, dia_max_slice_val = compute_volumne_frustum(min_areas_selection["diastole"], min_areas_selection[dist_col])
systole_vol, sys_max_slice_val = compute_volumne_frustum(min_areas_selection["systole"], min_areas_selection[dist_col])
low_conf_diastole_vol = round(compute_volumne_frustum(min_areas_selection["diastole_conf"], min_areas_selection[dist_col])[0], 2)
low_conf_systole_vol = round(compute_volumne_frustum(min_areas_selection["systole_conf"], min_areas_selection[dist_col])[0], 2)
return diastole_vol, systole_vol, low_conf_diastole_vol, low_conf_systole_vol, dia_frame, sys_frame, dia_max_slice_val, sys_max_slice_val
def count_pixels(patient_id, threshold, all_slice_data, model_name, threshold_value=-1):
patient_slice_data = all_slice_data[all_slice_data["patient_id"] == patient_id].copy()
patient_slice_data["slice_noloc"] = patient_slice_data["slice_no"].map(str) + "_" + patient_slice_data["slice_location"].map(str)
slices = patient_slice_data["slice_noloc"].unique().tolist()
frames = patient_slice_data["frame_no"].unique().tolist()
frame_count = len(frames)
# find lowest common set of frames that are present in every slice
for name, records in patient_slice_data.groupby("slice_noloc"):
slice_frames = records["frame_no"].unique()
if (len(frames) / 2) >= len(slice_frames):
# throw away small slices (416)
print "Patient " + str(patient_id) + ": throw away slice : " + str(name)
slices.remove(name)
continue
frames = list(set(frames) & set(slice_frames))
new_frame_count = len(frames)
if new_frame_count != frame_count:
print "Patient " + str(patient_id) + ": frames not the same for every slice : " + str(frame_count) + " <> " + str(new_frame_count)
file_name_slices = patient_slice_data.set_index('file_name')['slice_noloc'].to_dict()
file_name_frames = patient_slice_data.set_index('file_name')['frame_no'].to_dict()
# file_name_slice_locations = patient_slice_data.set_index('file_name')['frame_no'].to_dict()
# set up matrix indexed by slice_no and frame_no
slice_index = {}
for slice_no in slices:
slice_index[str(slice_no).rjust(2, '0')] = len(slice_index)
frame_pixel_series = {}
frame_confidence_series = {}
for frame_no in frames:
frame_pixel_series[str(frame_no).rjust(2, '0')] = [-1] * len(slices)
frame_confidence_series[str(frame_no).rjust(2, '0')] = [-1] * len(slices)
overlay_paths = helpers.get_patient_overlays(patient_id)
for overlay_path in overlay_paths:
overlay_img = cv2.imread(overlay_path, cv2.IMREAD_GRAYSCALE)
low_confidence_pixel_count = ((overlay_img < LOW_CONFIDENCE_PIXEL_THRESHOLD) & (overlay_img > 20)).sum()
# low_confidence_pixel_count = overlay_img[overlay_img > 10].mean()
if threshold_value >= 0:
overlay_img[overlay_img <= threshold_value] = 0
overlay_img[overlay_img > threshold_value] = 255
pixel_count = overlay_img.sum() / 255
#if pixel_count < 1:
# low_confidence_pixel_percentage = -1
#pixel_count = len(overlay_img[overlay_img > 0])
file_name = get_filename(overlay_path)
if file_name not in file_name_slices:
continue
slice_no = file_name_slices[file_name]
frame_no = file_name_frames[file_name]
slice_str = str(slice_no).rjust(2, '0')
frame_str = str(frame_no).rjust(2, '0')
if slice_str not in slice_index:
print "Patient " + str(patient_id) + " : slice " + slice_str + " skipped"
patient_slice_data = patient_slice_data[patient_slice_data["slice_noloc"] != slice_no]
continue
if frame_str not in frame_pixel_series:
print "Patient " + str(patient_id) + " : frame " + frame_str + " skipped"
patient_slice_data = patient_slice_data[patient_slice_data["frame_no"] != frame_no]
continue
frames_pixel_serie = frame_pixel_series[frame_str]
frame_confidence_serie = frame_confidence_series[frame_str]
serie_index = slice_index[slice_str]
frames_pixel_serie[serie_index] = pixel_count
frame_confidence_serie[serie_index] = low_confidence_pixel_count
data_frame = pandas.DataFrame()
patient_slice_data_frame1 = patient_slice_data[patient_slice_data["frame_no"] == 1]
data_frame["slice"] = slices
data_frame["slice_thickness"] = patient_slice_data_frame1["slice_thickness"].values
data_frame["slice_location"] = patient_slice_data_frame1["slice_location"].values
data_frame["slice_dist"] = (patient_slice_data_frame1["slice_location"].shift(-1) - patient_slice_data_frame1["slice_location"]).values
data_frame["slice_dist"].fillna(data_frame["slice_dist"].mean(), inplace=True)
data_frame["slice_location2a"] = patient_slice_data_frame1["slice_location2"].values
data_frame["slice_location2b"] = patient_slice_data_frame1["slice_location2"].shift(-1).values
data_frame["slice_dist2"] = data_frame.apply(lambda row: compute_distance(row["slice_location2a"], row["slice_location2b"]), axis=1)
data_frame["slice_dist2"].fillna(data_frame["slice_dist2"].mean(), inplace=True)
deltas = abs(abs(data_frame["slice_dist"]) - abs(data_frame["slice_dist2"])).sum()
if deltas > 1:
print "slice_dist != slice_dist2 (" + str(deltas) + ")"
data_frame["time"] = patient_slice_data_frame1["time"].values
for frame_no in frames:
frame_str = str(frame_no).rjust(2, '0')
interpolated_series = interpolate_series(frame_pixel_series[frame_str], frame_str)
data_frame["fr_" + frame_str] = interpolated_series
if SMOOTHEN_FRAMES:
for frame_no in frames:
prev_frame = frame_no - 1
if prev_frame < 1:
prev_frame = 30
next_frame = frame_no + 1
if next_frame > 30:
next_frame = 1
this_frame_col = "fr_" + str(frame_no).rjust(2, '0')
next_frame_col = "fr_" + str(next_frame).rjust(2, '0')
prev_frame_col = "fr_" + str(prev_frame).rjust(2, '0')
if this_frame_col in data_frame.columns and next_frame_col in data_frame.columns and prev_frame_col in data_frame.columns:
data_frame[this_frame_col] = (data_frame[this_frame_col] + data_frame[prev_frame_col] + data_frame[next_frame_col]) / 3
for frame_no in frames:
frame_str = str(frame_no).rjust(2, '0')
data_frame["co_" + frame_str] = frame_confidence_series[frame_str]
patient_dir = helpers.get_pred_patient_dir(patient_id)
data_frame.to_csv(patient_dir + "\\areas_" + model_name + ".csv", sep=";")
return data_frame
