Merge pull request #255 from Sichao25/restore

Restore contour.py

spateo/plotting/static/__init__.py

@@ -4,6 +4,7 @@
 from .align import multi_slices
 from .bbs import delaunay, polygon
 from .colorlabel import color_label
+from .contour import spatial_domains
 from .dotplot import dotplot
 from .geo import geo
 from .glm import glm_fit, glm_heatmap

spateo/plotting/static/contour.py

@@ -0,0 +1,68 @@
+"""Written by @Jinerhal, adapted by @Xiaojieqiu.
+"""
+from typing import Dict, Optional, Tuple, Union
+
+import cv2
+import numpy as np
+from anndata import AnnData
+
+from ...configuration import SKM
+from .utils import save_return_show_fig_utils
+
+
+@SKM.check_adata_is_type(SKM.ADATA_UMI_TYPE, "adata")
+def spatial_domains(
+    adata: AnnData,
+    bin_size: Optional[int] = None,
+    spatial_key: str = "spatial",
+    label_key: str = "cluster_img_label",
+    plot_size=(3, 3),
+    save_img="spatial_domains.png",
+):
+    """Generate an image with contours of each spatial domains.
+
+    Args:
+        adata: The adata object used to create the image.
+        bin_size: The size of the binning. Default to None.
+        spatial_key: The key name of the spatial coordinates. Default to "spatial".
+        label_key: The key name of the image label values. Default to "cluster_img_label".
+        plot_size: figsize for showing the image.
+        save_img: path to saving image file.
+    """
+    import matplotlib.pyplot as plt
+    from numpngw import write_png
+
+    label_list = np.unique(adata.obs[label_key])
+    labels = np.zeros(len(adata))
+    for i in range(len(label_list)):
+        labels[adata.obs[label_key] == label_list[i]] = i + 1
+
+    if bin_size is None:
+        bin_size = adata.uns["bin_size"]
+
+    label_img = np.zeros(
+        (
+            int(max(adata.obsm[spatial_key][:, 0] // bin_size)) + 1,
+            int(max(adata.obsm[spatial_key][:, 1] // bin_size)) + 1,
+        )
+    )
+    for i in range(len(adata)):
+        label_img[
+            int(adata.obsm[spatial_key][i, 0] // bin_size), int(adata.obsm[spatial_key][i, 1] // bin_size)
+        ] = labels[i]
+
+    contour_img = label_img.copy()
+    contour_img[:, :] = 255
+    for i in np.unique(label_img):
+        if i == 0:
+            continue
+        label_img_gray = np.where(label_img == i, 0, 1).astype("uint8")
+        _, thresh = cv2.threshold(label_img_gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
+        contour, _ = cv2.findContours(thresh, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_NONE)
+        contour_img = cv2.drawContours(contour_img, contour[:], -1, 0.5, 1)
+
+    fig = plt.figure()
+    fig.set_size_inches(plot_size[0], plot_size[1])
+    plt.imshow(contour_img, cmap="tab20", origin="lower")
+
+    write_png(save_img, contour_img.astype("uint8"))