Fluorescence channel 3D volume segmentation #213
Conversation
Applies on whole 3D stack. Image stack normalized based on center slice intensity range.
Soorya19Pradeep
requested review from
mattersoflight,
Christianfoley and
ziw-liu
micro_dl/utils/masks.py
Outdated
|
|
||
| elif style == 'binary': | ||
| ret, mask = cv2.threshold(input_image_norm.astype(np.uint8), thresh_input, 255, cv2.THRESH_BINARY) | ||
| input_image_norm = 255*(input_image_blur - thresh_input[1])/(thresh_input[0] - thresh_input[1]) |
There was a problem hiding this comment.
Is thresh_input an int or an iterable here? Can you add its parameter description to the docstring?
There was a problem hiding this comment.
Thank you @Christianfoley! I have added the docstring for thresh_input.
There was a problem hiding this comment.
Small thing, if the parameter is Iterable but the default is set to 0, the parameter type description should be "0 or Iterable". I would suggest in this case you set the default parameter to an empty list (description: "Iterable" makes sense) or None (description to "Iterable or None") according to Sphinx docstrings.
| focus_max = np.max(input_image_blur[mid_slice_id,:,:]) | ||
| focus_min = np.min(input_image_blur[mid_slice_id,:,:]) | ||
|
|
||
| input_image_norm = (input_image_blur - focus_min) / ( |
There was a problem hiding this comment.
Why is unit intensity needed if only the binary mask is returned?
There was a problem hiding this comment.
Do you mean the rescaling the intensity values of all slices based on intensity value range of the middle slice?
There was a problem hiding this comment.
Yes, since this operation does not skew the histogram, it should not affect otsu.
The otsu thresholding for fluorescence segmentation extended for 3D volume segmentation in preprocessing module. This avoids the deconvolution artifacts in out-of-plane slices. The normalization of slices was performed based on intensity range of center slice and simple thresholding was perform from the threshold value of center slice.