Skip to content

Commit

Permalink
clustering safeguards
Browse files Browse the repository at this point in the history
  • Loading branch information
@kapoorlab
kapoorlab committed Mar 2, 2024
1 parent 9eac706 commit 758541d
Show file tree
Hide file tree
Showing 2 changed files with 122 additions and 86 deletions.
204 changes: 120 additions & 84 deletions src/napatrackmater/Trackvector.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -32,12 +32,12 @@
class TrackVector(TrackMate):
def __init__(
self,
viewer: napari.Viewer,
image: np.ndarray,
master_xml_path: Path,
spot_csv_path: Path,
track_csv_path: Path,
edges_csv_path: Path,
viewer: napari.Viewer = None,
image: np.ndarray = None,
spot_csv_path: Path = None,
track_csv_path: Path = None,
edges_csv_path: Path = None,
t_minus: int = 0,
t_plus: int = 10,
x_start: int = 0,
Expand Down Expand Up @@ -358,7 +358,12 @@ def _interactive_function(self):
[r.result() for r in concurrent.futures.as_completed(futures)]

print("getting attributes")
self._get_attributes()
if (
self.spot_csv_path is not None
and self.track_csv_path is not None
and self.edges_csv_path is not None
):
self._get_attributes()
if self.autoencoder_model is not None:
self._compute_latent_space()
self.unique_tracks = {}
Expand Down Expand Up @@ -2298,6 +2303,9 @@ def core_clustering(
best_dynamic_cluster_labels = None
best_shape_cluster_labels = None
best_shape_dynamic_linkage_matrix = None
best_dynamic_linkage_matrix = None
best_shape_linkage_matrix = None
best_shape_dynamic_linkage_matrix = None

best_cluster_distance_map_shape_dynamic = None
best_cluster_distance_map_dynamic = None
Expand Down Expand Up @@ -2349,11 +2357,16 @@ def core_clustering(
shape_dynamic_linkage_matrix = linkage(
shape_dynamic_cosine_distance, method=method
)
shape_dynamic_cluster_labels = fcluster(
shape_dynamic_linkage_matrix,
cluster_threshold_shape_dynamic,
criterion=criterion,
)
try:
shape_dynamic_cluster_labels = fcluster(
shape_dynamic_linkage_matrix,
cluster_threshold_shape_dynamic,
criterion=criterion,
)
except Exception:
shape_dynamic_cluster_labels = fcluster(
shape_dynamic_linkage_matrix, 1, criterion="maxclust"
)

cluster_distance_map_shape_dynamic = calculate_intercluster_distance(
compute_vectors, shape_dynamic_cluster_labels
Expand All @@ -2363,54 +2376,64 @@ def core_clustering(
position_vector_1d, shape_dynamic_cluster_labels
)
)
try:

shape_dynamic_cluster_centroids = calculate_cluster_centroids(
shape_dynamic_eigenvectors_1d, shape_dynamic_cluster_labels
)

shape_dynamic_cluster_centroids = calculate_cluster_centroids(
shape_dynamic_eigenvectors_1d, shape_dynamic_cluster_labels
)
try:
shape_dynamic_silhouette = silhouette_score(
shape_dynamic_eigenvectors_1d,
shape_dynamic_cluster_labels,
metric=metric,
)
except Exception:
shape_dynamic_silhouette = 0
try:
shape_dynamic_wcss_value = calculate_wcss(
shape_dynamic_eigenvectors_1d,
shape_dynamic_cluster_labels,
shape_dynamic_cluster_centroids,
)
if use_sillhouette_criteria:
condition = shape_dynamic_silhouette > best_silhouette_shape_dynamic
else:
condition = shape_dynamic_wcss_value < best_wcss_shape_dynamic_value

if condition:
best_silhouette_shape_dynamic = shape_dynamic_silhouette
best_threshold_shape_dynamic = cluster_threshold_shape_dynamic
best_wcss_shape_dynamic_value = shape_dynamic_wcss_value
best_shape_dynamic_cluster_labels = shape_dynamic_cluster_labels
best_shape_dynamic_linkage_matrix = shape_dynamic_linkage_matrix
best_cluster_distance_map_shape_dynamic = (
cluster_distance_map_shape_dynamic
)
best_cluster_eucledian_distance_map_shape_dynamic = (
cluster_eucledian_distance_map_shape_dynamic
)
except Exception:
shape_dynamic_wcss_value = np.inf

if use_sillhouette_criteria:
condition = (shape_dynamic_silhouette > best_silhouette_shape_dynamic) & (
shape_dynamic_silhouette > 0
)
else:
condition = (shape_dynamic_wcss_value < best_wcss_shape_dynamic_value) & (
shape_dynamic_wcss_value > 0
)

shape_dynamic_silhouette = np.nan
shape_dynamic_wcss_value = np.nan
if condition:
best_silhouette_shape_dynamic = shape_dynamic_silhouette
best_threshold_shape_dynamic = cluster_threshold_shape_dynamic
best_wcss_shape_dynamic_value = shape_dynamic_wcss_value
best_shape_dynamic_cluster_labels = shape_dynamic_cluster_labels
best_shape_dynamic_linkage_matrix = shape_dynamic_linkage_matrix
best_cluster_distance_map_shape_dynamic = cluster_distance_map_shape_dynamic
best_cluster_eucledian_distance_map_shape_dynamic = (
cluster_eucledian_distance_map_shape_dynamic
)

for cluster_threshold_dynamic in cluster_threshold_dynamic_range:

dynamic_cosine_distance = pdist(dynamic_eigenvectors_1d, metric=metric)

dynamic_linkage_matrix = linkage(dynamic_cosine_distance, method=method)
dynamic_cluster_labels = fcluster(
dynamic_linkage_matrix,
cluster_threshold_dynamic,
criterion=criterion,
)
try:
dynamic_cluster_labels = fcluster(
dynamic_linkage_matrix,
cluster_threshold_dynamic,
criterion=criterion,
)
except Exception:
dynamic_cluster_labels = fcluster(
dynamic_linkage_matrix,
1,
criterion="maxclust",
)

cluster_distance_map_dynamic = calculate_intercluster_distance(
compute_vectors, dynamic_cluster_labels
Expand All @@ -2421,47 +2444,55 @@ def core_clustering(
)
)

dynamic_cluster_centroids = calculate_cluster_centroids(
dynamic_eigenvectors_1d, dynamic_cluster_labels
)
try:

dynamic_cluster_centroids = calculate_cluster_centroids(
dynamic_eigenvectors_1d, dynamic_cluster_labels
)

dynamic_silhouette = silhouette_score(
dynamic_eigenvectors_1d, dynamic_cluster_labels, metric=metric
)
except Exception:
dynamic_silhouette = 0
try:
dynamic_wcss_value = calculate_wcss(
dynamic_eigenvectors_1d,
dynamic_cluster_labels,
dynamic_cluster_centroids,
)
if use_sillhouette_criteria:
condition = dynamic_silhouette > best_silhouette_dynamic
else:
condition = dynamic_wcss_value < best_wcss_dynamic_value

if condition:
best_silhouette_dynamic = dynamic_silhouette
best_threshold_dynamic = cluster_threshold_dynamic
best_wcss_dynamic_value = dynamic_wcss_value
best_dynamic_cluster_labels = dynamic_cluster_labels
best_dynamic_linkage_matrix = dynamic_linkage_matrix
best_cluster_distance_map_dynamic = cluster_distance_map_dynamic
best_cluster_eucledian_distance_map_dynamic = (
cluster_eucledian_distance_map_dynamic
)

except Exception:
dynamic_wcss_value = np.inf
if use_sillhouette_criteria:
condition = (dynamic_silhouette > best_silhouette_dynamic) & (
dynamic_silhouette > 0
)
else:
condition = (dynamic_wcss_value < best_wcss_dynamic_value) & (
dynamic_wcss_value > 0
)

if condition:
best_silhouette_dynamic = dynamic_silhouette
best_threshold_dynamic = cluster_threshold_dynamic
best_wcss_dynamic_value = dynamic_wcss_value
best_dynamic_cluster_labels = dynamic_cluster_labels
best_dynamic_linkage_matrix = dynamic_linkage_matrix
best_cluster_distance_map_dynamic = cluster_distance_map_dynamic
best_cluster_eucledian_distance_map_dynamic = (
cluster_eucledian_distance_map_dynamic
)

dynamic_silhouette = np.nan
dynamic_wcss_value = np.nan
for cluster_threshold_shape in cluster_threshold_shape_range:
shape_cosine_distance = pdist(shape_eigenvectors_1d, metric=metric)

shape_linkage_matrix = linkage(shape_cosine_distance, method=method)
shape_cluster_labels = fcluster(
shape_linkage_matrix, cluster_threshold_shape, criterion=criterion
)
try:
shape_cluster_labels = fcluster(
shape_linkage_matrix, cluster_threshold_shape, criterion=criterion
)
except Exception:
shape_cluster_labels = fcluster(
shape_linkage_matrix, 1, criterion="maxclust"
)

shape_cluster_centroids = calculate_cluster_centroids(
shape_eigenvectors_1d, shape_cluster_labels
Expand All @@ -2479,29 +2510,34 @@ def core_clustering(
shape_silhouette = silhouette_score(
shape_eigenvectors_1d, shape_cluster_labels, metric=metric
)
except Exception:
shape_silhouette = 0
try:
shape_wcss_value = calculate_wcss(
shape_eigenvectors_1d, shape_cluster_labels, shape_cluster_centroids
)
if use_sillhouette_criteria:
condition = shape_silhouette > best_silhouette_shape
else:
condition = shape_wcss_value < best_wcss_shape_value

if condition:
best_silhouette_shape = shape_silhouette
best_threshold_shape = cluster_threshold_shape
best_wcss_shape_value = shape_wcss_value
best_shape_cluster_labels = shape_cluster_labels
best_shape_linkage_matrix = shape_linkage_matrix
best_cluster_distance_map_shape = cluster_distance_map_shape
best_cluster_eucledian_distance_map_shape = (
cluster_eucledian_distance_map_shape
)

except Exception:
shape_wcss_value = np.inf
if use_sillhouette_criteria:
condition = (shape_silhouette > best_silhouette_shape) & (
shape_silhouette > 0
)
else:
condition = (shape_wcss_value < best_wcss_shape_value) & (
shape_wcss_value > 0
)

if condition:
best_silhouette_shape = shape_silhouette
best_threshold_shape = cluster_threshold_shape
best_wcss_shape_value = shape_wcss_value
best_shape_cluster_labels = shape_cluster_labels
best_shape_linkage_matrix = shape_linkage_matrix
best_cluster_distance_map_shape = cluster_distance_map_shape
best_cluster_eucledian_distance_map_shape = (
cluster_eucledian_distance_map_shape
)

shape_silhouette = np.nan
shape_wcss_value = np.nan
print(
f"best threshold value for shape dynamic {best_threshold_shape_dynamic} with silhouette score of {best_silhouette_shape_dynamic} and with wcss score of {best_wcss_shape_dynamic_value}, number of clusters {len(np.unique(best_shape_dynamic_cluster_labels))}"
)
Expand Down
4 changes: 2 additions & 2 deletions src/napatrackmater/_version.py
View file
Original file line number Diff line number Diff line change
@@ -1,2 +1,2 @@
__version__ = version = "4.9.0"
__version_tuple__ = version_tuple = (4, 9, 0)
__version__ = version = "4.9.1"
__version_tuple__ = version_tuple = (4, 9, 1)

0 comments on commit 758541d

Please sign in to comment.