Skip to content

Commit

Permalink
compute raw matrix
Browse files Browse the repository at this point in the history
  • Loading branch information
@kapoorlab
kapoorlab committed Jan 23, 2024
1 parent 4149f4b commit 2e7b3ee
Show file tree
Hide file tree
Showing 2 changed files with 184 additions and 0 deletions.
182 changes: 182 additions & 0 deletions src/napatrackmater/Trackvector.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1398,6 +1398,171 @@ def calculate_cluster_centroids(data, labels):
centroids.append(centroid)
return np.array(centroids)

def simple_unsupervised_clustering(
full_dataframe,
csv_file_name,
analysis_vectors,
cluster_threshold=3,
metric="euclidean",
method="centroid",
criterion="distance",
):


csv_file_name_original = csv_file_name
analysis_track_ids = []
shape_dynamic_covariance_matrix = []
shape_covariance_matrix = []
dynamic_covariance_matrix = []
for track_id, (
shape_dynamic_dataframe_list,
shape_dataframe_list,
dynamic_dataframe_list,
full_dataframe_list,
) in analysis_vectors.items():
shape_dynamic_track_array = np.array(
[
[item for item in record.values()]
for record in shape_dynamic_dataframe_list
]
)
shape_track_array = np.array(
[[item for item in record.values()] for record in shape_dataframe_list]
)
dynamic_track_array = np.array(
[[item for item in record.values()] for record in dynamic_dataframe_list]
)
assert (
shape_dynamic_track_array.shape[0]
== shape_track_array.shape[0]
== dynamic_track_array.shape[0]
), "Shape dynamic, shape and dynamic track arrays must have the same length."
if shape_dynamic_track_array.shape[0] > 1:

covariance_computation_shape_dynamic = compute_raw_matrix(shape_dynamic_track_array)

covaraince_computation_shape = compute_raw_matrix(
shape_track_array
)

covaraince_computation_dynamic = compute_raw_matrix(
dynamic_track_array
)
if covariance_computation_shape_dynamic is not None and covaraince_computation_shape is not None and covaraince_computation_dynamic is not None:

shape_dynamic_eigenvectors = covariance_computation_shape_dynamic
shape_eigenvectors = covaraince_computation_shape
dynamic_eigenvectors = covaraince_computation_dynamic
shape_dynamic_covariance_matrix.append(shape_dynamic_eigenvectors)
shape_covariance_matrix.append(shape_eigenvectors)
dynamic_covariance_matrix.append(dynamic_eigenvectors)
analysis_track_ids.append(track_id)
if len(shape_dynamic_covariance_matrix) > 0 and len(shape_covariance_matrix) > 0 and len(dynamic_covariance_matrix) > 0:
shape_dynamic_covariance_3d = np.dstack(shape_dynamic_covariance_matrix)
shape_covariance_3d = np.dstack(shape_covariance_matrix)
dynamic_covariance_3d = np.dstack(dynamic_covariance_matrix)

shape_dynamic_covariance_matrix = np.mean(shape_dynamic_covariance_matrix, axis=0)
shape_covariance_matrix = np.mean(shape_covariance_matrix, axis=0)
dynamic_covariance_matrix = np.mean(dynamic_covariance_matrix, axis=0)

shape_dynamic_covariance_2d = shape_dynamic_covariance_3d.reshape(
len(analysis_track_ids), -1
)
shape_covariance_2d = shape_covariance_3d.reshape(len(analysis_track_ids), -1)
dynamic_covariance_2d = dynamic_covariance_3d.reshape(len(analysis_track_ids), -1)

track_arrays_array = [
shape_dynamic_covariance_matrix,
shape_covariance_matrix,
dynamic_covariance_matrix,
]

track_arrays_array_names = ["shape_dynamic", "shape", "dynamic"]
clusterable_track_arrays = [
shape_dynamic_covariance_2d,
shape_covariance_2d,
dynamic_covariance_2d,
]

for track_arrays in track_arrays_array:
clusterable_track_array = clusterable_track_arrays[
track_arrays_array.index(track_arrays)
]
shape_dynamic_cosine_distance = pdist(clusterable_track_array, metric=metric)

shape_dynamic_linkage_matrix = linkage(
shape_dynamic_cosine_distance, method=method
)
shape_dynamic_cluster_labels = fcluster(
shape_dynamic_linkage_matrix, cluster_threshold, criterion=criterion
)

cluster_centroids = calculate_cluster_centroids(
clusterable_track_array, shape_dynamic_cluster_labels
)
silhouette = silhouette_score(
clusterable_track_array, shape_dynamic_cluster_labels, metric=metric
)
wcss_value = calculate_wcss(
clusterable_track_array, shape_dynamic_cluster_labels, cluster_centroids
)

silhouette_file_name = os.path.join(
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ f"_silhouette_{metric}_{cluster_threshold}.npy"
)
np.save(silhouette_file_name, silhouette)

wcss_file_name = os.path.join(
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ f"_wcss_{metric}_{cluster_threshold}.npy"
)
np.save(wcss_file_name, wcss_value)
track_id_to_cluster = {
track_id: cluster_label
for track_id, cluster_label in zip(
analysis_track_ids, shape_dynamic_cluster_labels
)
}
full_dataframe["Cluster"] = full_dataframe["Track ID"].map(track_id_to_cluster)
result_dataframe = full_dataframe[["Track ID", "t", "z", "y", "x", "Cluster"]]
csv_file_name = (
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ ".csv"
)

if os.path.exists(csv_file_name):
os.remove(csv_file_name)
result_dataframe.to_csv(csv_file_name, index=False)

mean_matrix_file_name = (
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ f"_{metric}_covariance.npy"
)
np.save(mean_matrix_file_name, track_arrays)

linkage_npy_file_name = (
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ f"_{metric}_linkage.npy"
)
np.save(linkage_npy_file_name, shape_dynamic_linkage_matrix)

cluster_labels_npy_file_name = (
csv_file_name_original
+ track_arrays_array_names[track_arrays_array.index(track_arrays)]
+ f"_{metric}_cluster_labels.npy"
)
np.save(cluster_labels_npy_file_name, shape_dynamic_cluster_labels)





def unsupervised_clustering(
full_dataframe,
Expand Down Expand Up @@ -1771,6 +1936,23 @@ def convert_tracks_to_arrays(



def compute_raw_matrix(track_arrays, delta_t):

track_duration = track_arrays.shape[0]
delta_t = int(delta_t)
if track_duration < delta_t:
#zero pad
pad_rows = delta_t - track_duration

# Pad with the last row and then zeros
last_row = track_arrays[-1, :]
zero_pad = np.zeros((pad_rows, track_arrays.shape[1]))
track_arrays = np.vstack((track_arrays, np.tile(last_row, (pad_rows, 1)), zero_pad))

flattened_array = track_arrays.flatten()

return flattened_array


def compute_covariance_matrix(track_arrays):
try:
Expand Down
2 changes: 2 additions & 0 deletions src/napatrackmater/__init__.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
TrackVector,
convert_tracks_to_arrays,
unsupervised_clustering,
simple_unsupervised_clustering,
supervised_clustering,
create_global_gt_dataframe,
create_gt_analysis_vectors_dict,
Expand Down Expand Up @@ -73,6 +74,7 @@ def load_json(fpath):
"predict_supervised_clustering",
"convert_tracks_to_arrays",
"unsupervised_clustering",
"simple_unsupervised_clustering",
"load_json",
"create_mitosis_training_data",
"train_mitosis_classifier",
Expand Down

0 comments on commit 2e7b3ee

Please sign in to comment.