Cell segmentation

Functions used to segment cells.

Apply watershed algorithm

Main function to segment cells with a watershed algorithm:

Our segmentation using watershed algorithm can also be perform with two separated steps:

bigfish.segmentation.cell_watershed(image, nuc_label, threshold, alpha=0.8)

Apply watershed algorithm to segment cell instances.

In a watershed algorithm we consider cells as watershed to be flooded. The watershed relief is inversely proportional to both the pixel intensity and the closeness to nuclei. Pixels with a high intensity or close to labelled nuclei have a low watershed relief value. They will be flooded in priority. Flooding the watersheds allows to propagate nuclei labels through potential cytoplasm areas. The lines separating watershed are the final segmentation of the cells.

Parameters:
imagenp.ndarray, np.uint

Cells image with shape (y, x).

nuc_labelnp.ndarray, np.int64

Result of the nuclei segmentation with shape (y, x) and nuclei instances labelled.

thresholdint or float

Threshold to discriminate cells surfaces from background.

alphafloat or int

Weight of the pixel intensity values to compute the watershed relief.

Returns:
cell_labelnp.ndarray, np.int64

Segmentation of cells with shape (y, x).

bigfish.segmentation.get_watershed_relief(image, nuc_label, alpha)

Build a representation of cells as watershed.

In a watershed algorithm we consider cells as watershed to be flooded. The watershed relief is inversely proportional to both the pixel intensity and the closeness to nuclei. Pixels with a high intensity or close to labelled nuclei have a low watershed relief value. They will be flooded in priority. Flooding the watersheds allows to propagate nuclei labels through potential cytoplasm areas. The lines separating watershed are the final segmentation of the cells.

Parameters:
imagenp.ndarray, np.uint

Cells image with shape (z, y, x) or (y, x).

nuc_labelnp.ndarray, np.int64

Result of the nuclei segmentation with shape (y, x) and nuclei instances labelled.

alphafloat or int

Weight of the pixel intensity values to compute the relief.

Returns:
watershed_reliefnp.ndarray, np.uint16

Watershed representation of cells with shape (y, x).

bigfish.segmentation.apply_watershed(watershed_relief, nuc_label, cell_mask)

Apply watershed algorithm to segment cell instances.

In a watershed algorithm we consider cells as watershed to be flooded. The watershed relief is inversely proportional to both the pixel intensity and the closeness to nuclei. Pixels with a high intensity or close to labelled nuclei have a low watershed relief value. They will be flooded in priority. Flooding the watersheds allows to propagate nuclei labels through potential cytoplasm areas. The lines separating watershed are the final segmentation of the cells.

Parameters:
watershed_reliefnp.ndarray, np.uint or np.int

Watershed representation of cells with shape (y, x).

nuc_labelnp.ndarray, np.int64

Result of the nuclei segmentation with shape (y, x) and nuclei instances labelled.

cell_masknp.ndarray, bool

Binary image of cells surface with shape (y, x).

Returns:
cell_labelnp.ndarray, np.int64

Segmentation of cells with shape (y, x).

Apply a Unet-based model (distance map)

Load a pretrained model:

Segment cells:

See an example of application here.

bigfish.segmentation.unet_distance_edge_double()

Load a pretrained Unet model to predict foreground and a distance map to edge from nucleus and cell images.

Returns:
modeltensorflow.keras.model object

Pretrained Unet model.

bigfish.segmentation.apply_unet_distance_double(model, nuc, cell, nuc_label, target_size=None, test_time_augmentation=False)

Segment cell with a pretrained model to predict distance map and use it with a watershed algorithm.

Parameters:
modeltensorflow.keras.model object

Pretrained Unet model that predict distance to edges and cell surface.

nucnp.ndarray, np.uint

Original nucleus image with shape (y, x).

cellnp.ndarray, np.uint

Original cell image to segment with shape (y, x).

nuc_labelnp.ndarray, np.int64

Labelled nucleus image. Each nucleus is characterized by the same pixel value.

target_sizeint

Resize image before segmentation. A squared image is resize to target_size. A rectangular image is resize such that its smaller dimension equals target_size.

test_time_augmentationbool

Apply test time augmentation or not. The image is augmented 8 times and the final segmentation is the average result over these augmentations.

Returns:
cell_label_prednp.ndarray, np.int64

Labelled cell image. Each cell is characterized by the same pixel value.

bigfish.segmentation.from_distance_to_instances(label_x_nuc, label_2_cell, label_distance, nuc_3_classes=False, compute_nuc_label=False)

Extract instance labels from a distance map and a binary surface prediction with a watershed algorithm.

Parameters:
label_x_nucnp.ndarray, np.float32

Model prediction about the nucleus surface (and boundaries), with shape (y, x, 1) or (y, x, 3).

label_2_cellnp.ndarray, np.float32

Model prediction about cell surface, with shape (y, x, 1).

label_distancenp.ndarray, np.uint16

Model prediction about the distance to edges, with shape (y, x, 1).

nuc_3_classesbool

Nucleus image input is an output from a 3-classes Unet.

compute_nuc_labelbool

Extract nucleus instance labels.

Returns:
nuc_labelnp.ndarray, np.int64

Labelled nucleus image. Each nucleus is characterized by the same pixel value.

cell_labelnp.ndarray, np.int64

Labelled cell image. Each cell is characterized by the same pixel value.