Ma Nan
63c593f09a
fix: import fail if label are all zero fix: add cutter name to node fix: cannot append node lib file on NAS (UNC path) fix: join component cause disappear feat: smooth label feat: better topo on cutout by range
153 lines
4.7 KiB
Python
153 lines
4.7 KiB
Python
import copy
|
|
import numpy as np
|
|
|
|
from . import scipy
|
|
from . import scipy as ndi
|
|
|
|
# 3rd-party
|
|
import transforms3d
|
|
# TODO: turn to dataclasses
|
|
|
|
|
|
class Layer():
|
|
def __init__(self,
|
|
data: np.ndarray,
|
|
name: str,
|
|
kind="scalar",
|
|
affine=np.identity(4)) -> None:
|
|
if data.ndim != 5:
|
|
raise Exception("Data shape order should be TXYZC")
|
|
|
|
affine = np.array(affine)
|
|
if affine.shape != (4, 4):
|
|
raise Exception("affine shape should be (4,4)")
|
|
|
|
self.data = data
|
|
self.name = name
|
|
self.kind = kind
|
|
self.affine = affine
|
|
|
|
@property
|
|
def bioxel_size(self):
|
|
t, r, z, s = transforms3d.affines.decompose44(self.affine)
|
|
return z.tolist()
|
|
|
|
@property
|
|
def shape(self):
|
|
return self.data.shape[1:4]
|
|
|
|
@property
|
|
def dtype(self):
|
|
return self.data.dtype
|
|
|
|
@property
|
|
def origin(self):
|
|
t, r, z, s = transforms3d.affines.decompose44(self.affine)
|
|
return t.tolist()
|
|
|
|
@property
|
|
def euler(self):
|
|
t, r, z, s = transforms3d.affines.decompose44(self.affine)
|
|
return list(transforms3d.euler.mat2euler(r))
|
|
|
|
@property
|
|
def min(self):
|
|
return float(np.min(self.data))
|
|
|
|
@property
|
|
def frame_count(self):
|
|
return self.data.shape[0]
|
|
|
|
@property
|
|
def channel_count(self):
|
|
return self.data.shape[-1]
|
|
|
|
@property
|
|
def max(self):
|
|
return float(np.max(self.data))
|
|
|
|
def copy(self):
|
|
return copy.deepcopy(self)
|
|
|
|
def fill(self, value: float, mask: np.ndarray, smooth: int = 0):
|
|
mask_frames = ()
|
|
if mask.ndim == 4:
|
|
if mask.shape[0] != self.frame_count:
|
|
raise Exception("Mask frame count is not same as ")
|
|
for f in range(self.frame_count):
|
|
mask_frame = mask[f, :, :, :]
|
|
if smooth > 0:
|
|
mask_frame = scipy.median_filter(mask_frame.astype(np.float32),
|
|
mode="nearest",
|
|
size=smooth)
|
|
# mask_frame = scipy.median_filter(
|
|
# mask_frame.astype(np.float32), size=2)
|
|
mask_frames += (mask_frame,)
|
|
elif mask.ndim == 3:
|
|
for f in range(self.frame_count):
|
|
mask_frame = mask[:, :, :]
|
|
if smooth > 0:
|
|
mask_frame = scipy.median_filter(mask_frame.astype(np.float32),
|
|
mode="nearest",
|
|
size=smooth)
|
|
# mask_frame = scipy.median_filter(
|
|
# mask_frame.astype(np.float32), size=2)
|
|
mask_frames += (mask_frame,)
|
|
else:
|
|
raise Exception("Mask shape order should be TXYZ or XYZ")
|
|
|
|
_mask = np.stack(mask_frames)
|
|
_mask = np.expand_dims(_mask, axis=-1)
|
|
self.data = _mask * value + (1-_mask) * self.data
|
|
|
|
def resize(self, shape: tuple, smooth: int = 0, progress_callback=None):
|
|
if len(shape) != 3:
|
|
raise Exception("Shape must be 3 dim")
|
|
|
|
data = self.data
|
|
|
|
# # TXYZC > TXYZ
|
|
# if self.kind in ['label', 'scalar']:
|
|
# data = np.amax(data, -1)
|
|
|
|
# if self.kind in ['scalar']:
|
|
# dtype = data.dtype
|
|
# data = data.astype(np.float32)
|
|
|
|
data_frames = ()
|
|
for f in range(self.frame_count):
|
|
if progress_callback:
|
|
progress_callback(f, self.frame_count)
|
|
|
|
frame = data[f, :, :, :, :]
|
|
if smooth > 0:
|
|
frame = scipy.median_filter(frame.astype(np.float32),
|
|
mode="nearest",
|
|
size=smooth)
|
|
|
|
factors = np.divide(self.shape, shape)
|
|
zoom_factors = [1 / f for f in factors]
|
|
order = 0 if frame.dtype == bool else 1
|
|
frame = ndi.zoom(frame,
|
|
zoom_factors+[1.0],
|
|
mode="nearest",
|
|
grid_mode=False,
|
|
order=order)
|
|
if smooth > 0:
|
|
frame = frame.astype(self.dtype)
|
|
data_frames += (frame,)
|
|
|
|
data = np.stack(data_frames)
|
|
|
|
# if self.kind in ['scalar']:
|
|
# data = data.astype(dtype)
|
|
|
|
# TXYZ > TXYZC
|
|
# if self.kind in ['label', 'scalar']:
|
|
# data = np.expand_dims(data, axis=-1) # expend channel
|
|
|
|
self.data = data
|
|
|
|
mat_scale = transforms3d.zooms.zfdir2aff(factors[0])
|
|
self.affine = np.dot(self.affine, mat_scale)
|