def do_sampling(pat_df, out_dir):
for pat,side,view in pat_df.index.unique():
full_fn = const_filename(pat, side, view, full_img_dir)
# import pdb; pdb.set_trace()
try:
if target_width is None:
full_img = read_resize_img(
full_fn, target_height=target_height)
else:
full_img = read_resize_img(
full_fn, target_size=(target_height, target_width))
img_id = '_'.join([pat, side, view])
print "ID:%s, read image of size=%s" % (img_id, full_img.shape),
if segment_breast:
full_img, bbox = imprep.segment_breast(full_img)
print "size after segmentation=%s" % (str(full_img.shape))
sys.stdout.flush()
# Read mask image(s).
abn_path = roi_mask_path_df.loc[pat].loc[side].loc[view]
if isinstance(abn_path, pd.Series):
abn_num = [abn_path['abn_num']]
pathology = [abn_path['pathology']]
itypes = [abn_path['type']]
else:
abn_num = abn_path['abn_num']
pathology = abn_path['pathology']
itypes = abn_path['type']
bkg_sampled = False
for abn, path, itype in zip(abn_num, pathology, itypes):
mask_fn = const_filename(pat, side, view, roi_mask_dir, itype, abn)
if target_width is None:
mask_img = read_resize_img(
mask_fn, target_height=target_height, gs_255=True)
else:
mask_img = read_resize_img(
mask_fn, target_size=(target_height, target_width),
gs_255=True)
if segment_breast:
mask_img = crop_img(mask_img, bbox)
# sample using mask and full image.
nb_hns_ = nb_hns if not bkg_sampled else 0
if nb_hns_ > 0:
hns_sampled = sample_blob_negatives(
full_img, mask_img, out_dir, img_id,
abn, blob_detector, patch_size, neg_cutoff,
nb_hns_, 0, bkg_dir, verbose)
else:
hns_sampled = 0
pos = path.startswith('MALIGNANT')
nb_bkg_ = nb_bkg - hns_sampled if not bkg_sampled else 0
sample_patches(full_img, mask_img, out_dir, img_id, abn, pos,
patch_size, pos_cutoff, neg_cutoff,
nb_bkg_, nb_abn, hns_sampled, itype,
bkg_dir, calc_pos_dir, calc_neg_dir,
mass_pos_dir, mass_neg_dir, verbose)
bkg_sampled = True
except AttributeError:
print "Read image error: %s" % (full_fn)
except ValueError:
print "Error sampling from ROI mask image: %s" % (mask_fn)
def do_sampling(pat_df, out_dir):
for pat, side, view in pat_df.index.unique():
full_fn = const_filename(pat, side, view, full_img_dir)
# import pdb; pdb.set_trace()
try:
full_img = read_resize_img(full_fn,
target_height=target_height)
img_id = '_'.join([pat, side, view])
print "ID:%s, read image of size=%s" % (img_id,
full_img.shape),
full_img, bbox = imprep.segment_breast(full_img)
print "size after segmentation=%s" % (str(full_img.shape))
sys.stdout.flush()
# Read mask image(s).
abn_path = roi_mask_path_df.loc[pat].loc[side].loc[view]
if isinstance(abn_path, pd.Series):
abn_num = [abn_path['abn_num']]
pathology = [abn_path['pathology']]
itypes = [abn_path['type']]
else:
abn_num = abn_path['abn_num']
pathology = abn_path['pathology']
itypes = abn_path['type']
bkg_sampled = False
for abn, path, itype in zip(abn_num, pathology, itypes):
mask_fn = const_filename(pat, side, view, roi_mask_dir,
itype, abn)
mask_img = read_resize_img(mask_fn,
target_height=target_height,
gs_255=True)
mask_img = crop_img(mask_img, bbox)
# sample using mask and full image.
nb_hns_ = nb_hns if not bkg_sampled else 0
if nb_hns_ > 0:
hns_sampled = sample_blob_negatives(
full_img, mask_img, out_dir, img_id, abn,
blob_detector, patch_size, neg_cutoff, nb_hns_, 0,
bkg_dir, verbose)
else:
hns_sampled = 0
pos = path.startswith('MALIGNANT')
nb_bkg_ = nb_bkg - hns_sampled if not bkg_sampled else 0
sample_patches(full_img, mask_img, out_dir, img_id, abn,
pos, patch_size, pos_cutoff, neg_cutoff,
nb_bkg_, nb_abn, hns_sampled, itype,
bkg_dir, calc_pos_dir, calc_neg_dir,
mass_pos_dir, mass_neg_dir, verbose)
bkg_sampled = True
except AttributeError:
print "Read image error: %s" % (full_fn)
except ValueError:
print "Error sampling from ROI mask image: %s" % (mask_fn)
def score_write_patches(img_list,
lab_list,
target_height,
target_scale,
patch_size,
stride,
model,
batch_size,
neg_out,
pos_out,
bkg_out,
preprocess=None,
equalize_hist=False,
featurewise_center=False,
featurewise_mean=91.6,
roi_cutoff=.9,
bkg_cutoff=[.5, 1.],
sample_bkg=True,
img_ext='png',
random_seed=12345,
parallelized=False):
'''Score image patches and write them to an external directory
'''
def write_patches(img_fn, patch_dat, idx, out_dir, img_ext='png'):
basename = os.path.basename(img_fn)
fn_no_ext = os.path.splitext(basename)[0]
if img_ext == 'png':
max_val = 65535.
else:
max_val = 255.
for i in idx:
patch = patch_dat[i]
patch_max = patch.max() if patch.max() != 0 else max_val
patch *= max_val / patch_max
patch = patch.astype('int32')
mode = 'I' if img_ext == 'png' else None
patch_img = toimage(patch,
high=patch.max(),
low=patch.min(),
mode=mode)
filename = fn_no_ext + "_%06d" % (i) + '.' + img_ext
fullname = os.path.join(out_dir, filename)
patch_img.save(fullname)
rng = RandomState(random_seed)
nb_roi = 0
nb_bkg = 0
for img_fn, img_lab in zip(img_list, lab_list):
img = read_resize_img(img_fn, target_height=target_height)
img, _ = prep.segment_breast(img)
img = add_img_margins(img, patch_size / 2)
patch_dat, nb_row, nb_col = sweep_img_patches(
img,
patch_size,
stride,
target_scale=target_scale,
equalize_hist=equalize_hist)
org_patch_dat = patch_dat.copy()
if parallelized and len(patch_dat) % 2 == 1:
last_patch = patch_dat[-1:, :, :]
patch_dat = np.append(patch_dat, last_patch, axis=0)
appended = True
else:
appended = False
if dim_ordering == 'th':
patch_X = np.zeros((patch_dat.shape[0], 3, patch_dat.shape[1],
patch_dat.shape[2]),
dtype='float64')
patch_X[:, 0, :, :] = patch_dat
patch_X[:, 1, :, :] = patch_dat
patch_X[:, 2, :, :] = patch_dat
else:
patch_X = np.zeros((patch_dat.shape[0], patch_dat.shape[1],
patch_dat.shape[2], 3),
dtype='float64')
patch_X[:, :, :, 0] = patch_dat
patch_X[:, :, :, 1] = patch_dat
patch_X[:, :, :, 2] = patch_dat
if featurewise_center:
patch_X -= featurewise_mean
elif preprocess is not None:
patch_X = preprocess(patch_X)
pred = model.predict(patch_X, batch_size=batch_size)
# import pdb; pdb.set_trace()
if appended:
pred = pred[:-1]
roi_idx = np.where(pred[:, 0] < 1 - roi_cutoff)[0]
bkg_idx = np.where(
np.logical_and(pred[:, 0] > bkg_cutoff[0],
pred[:, 0] <= bkg_cutoff[1]))[0]
if sample_bkg and len(bkg_idx) > len(roi_idx):
bkg_idx = rng.choice(bkg_idx, len(roi_idx), replace=False)
roi_out = pos_out if img_lab == 1 else neg_out
write_patches(img_fn, org_patch_dat, roi_idx, roi_out, img_ext)
write_patches(img_fn, org_patch_dat, bkg_idx, bkg_out, img_ext)
nb_roi += len(roi_idx)
nb_bkg += len(bkg_idx)
return nb_roi, nb_bkg
def do_sampling(pat_df, out_dir):
#遍历病人信息
for pat, side, view in pat_df.index.unique():
#文件名命名
full_fn = const_filename(pat, side, view, full_img_dir)
# import pdb; pdb.set_trace()
try:
#如果未指定宽度则根据高度调整图片大小
if target_width is None:
full_img = read_resize_img(full_fn,
target_height=target_height)
else:
#根据宽高调整大小
full_img = read_resize_img(full_fn,
target_size=(target_height,
target_width))
#命名图片ID
img_id = '_'.join([pat, side, view])
print(
"ID:%s, read image of size=%s" %
(img_id, full_img.shape), )
#图像中分割出乳腺为true
if segment_breast:
#获取乳腺图像和坐标宽高
full_img, bbox = imprep.segment_breast(full_img)
#输出图像宽高通道数
print("size after segmentation=%s" % (str(full_img.shape)))
sys.stdout.flush()
# Read mask image(s).读取掩码图
#获取异常路径
abn_path = roi_mask_path_df.loc[pat].loc[side].loc[view]
#无论此路径是否为pd.Series类型，都要进行异常数，病理（良恶性），类型（肿块或钙化点）的数据获取
if isinstance(abn_path, pd.Series):
abn_num = [abn_path['abn_num']]
pathology = [abn_path['pathology']]
itypes = [abn_path['type']]
else:
abn_num = abn_path['abn_num']
pathology = abn_path['pathology']
itypes = abn_path['type']
#
bkg_sampled = False
#遍历刚才获取的数据，转为掩码文件
for abn, path, itype in zip(abn_num, pathology, itypes):
mask_fn = const_filename(pat, side, view, roi_mask_dir,
itype, abn)
if target_width is None:
mask_img = read_resize_img(mask_fn,
target_height=target_height,
gs_255=True)
else:
mask_img = read_resize_img(mask_fn,
target_size=(target_height,
target_width),
gs_255=True)
if segment_breast:
mask_img = crop_img(mask_img, bbox)
# sample using mask and full image.样本使用掩码和全图
#获取难类数
nb_hns_ = nb_hns if not bkg_sampled else 0
if nb_hns_ > 0:
#返回阴性斑点采样得到的背景样本数
hns_sampled = sample_blob_negatives(
full_img, mask_img, out_dir, img_id, abn,
blob_detector, patch_size, neg_cutoff, nb_hns_, 0,
bkg_dir, verbose)
else:
hns_sampled = 0
pos = path.startswith('MALIGNANT')
#背景数为总背景数-阴性斑点采样数
nb_bkg_ = nb_bkg - hns_sampled if not bkg_sampled else 0
#对块进行采样
sample_patches(full_img, mask_img, out_dir, img_id, abn,
pos, patch_size, pos_cutoff, neg_cutoff,
nb_bkg_, nb_abn, hns_sampled, itype,
bkg_dir, calc_pos_dir, calc_neg_dir,
mass_pos_dir, mass_neg_dir, verbose)
bkg_sampled = True
except AttributeError:
print("Read image error: %s" % (full_fn))
except ValueError:
print("Error sampling from ROI mask image: %s" % (mask_fn))
def get_image_and_mask(split, patient_id, side, view, image_dir, roi_mask_dir,
abn_type, abn_id, target_height, target_width):
token_list = []
token_list.append(("Calc" if abn_type == "calcification" else "Mass") +
"-" + split)
token_list.extend([patient_id, side, view])
# get image directory
image_folder_name = "_".join(token_list)
if split == "Training":
image_dir = os.path.join(image_dir, image_folder_name)
else:
image_dir = os.path.join(image_dir, "CBIS-DDSM", image_folder_name)
# search for the image
image_path = None
for cur_dir, sub_folders, file_names in os.walk(image_dir):
if file_names:
for file_name in file_names:
if file_name.endswith(".png"):
assert file_name == "000000.png"
image_path = os.path.join(cur_dir, "000000.png")
break
if not image_path:
raise RuntimeError(
"Can not find image file in {!r}.".format(image_dir))
ori_image_shape = cv2.imread(image_path).shape[:2]
if target_width is None:
full_image = read_resize_img(image_path, target_height=target_height)
else:
full_image = read_resize_img(image_path,
target_size=(target_height, target_width))
# get ROI mask directory
# NOTE ROI and mask are stored in one folder, and 0 and 1 does not mean ROI
# and mask, may be revsered!
token_list.append(str(abn_id))
roi_mask_folder_name = "_".join(token_list)
roi_mask_paths = []
if split == "Training":
roi_mask_dir = os.path.join(roi_mask_dir, roi_mask_folder_name)
else:
roi_mask_dir = os.path.join(roi_mask_dir, "CBIS-DDSM",
roi_mask_folder_name)
for cur_dir, sub_folders, file_names in os.walk(roi_mask_dir):
if file_names:
for file_name in file_names:
if file_name.endswith(".png"):
assert file_name in ("000000.png", "000001.png")
roi_mask_paths.append(os.path.join(cur_dir, file_name))
if len(roi_mask_paths) != 2:
raise RuntimeError("Wrong number of image files in %s." % roi_mask_dir)
mask_path = None
for path in roi_mask_paths:
mask = cv2.imread(path)
if mask.shape[:2] == ori_image_shape:
mask_path = path
break
if mask_path is None:
raise RuntimeError("Found no mask image with a compatiable shape.")
if target_width is None:
mask_image = read_resize_img(mask_path,
target_height=target_height,
gs_255=True)
else:
mask_image = read_resize_img(mask_path,
target_size=(target_height, target_width),
gs_255=True)
return full_image, mask_image