Update subtile figure

src/figures/biopsy.py

@@ -1,13 +1,79 @@
 import pandas as pd
 from pathlib import Path
+import matplotlib.pyplot as plt
+import seaborn as sns
 
-
-df = pd.read_csv(Path("data", "bxs", "9_2_1.csv"))
+patient = "9_2"
+df = pd.read_csv(Path("data", "bxs", f"{patient}_1.csv"))
+protein = "pRB"
 
 # visualize the biopsy based on the spatial coord X_centroid and Y_centroid
+def select_non_overlapping_tiles(df, num_tiles=20):
+    selected_tiles = []
+    for _, row in df.iterrows():
+        if len(selected_tiles) >= num_tiles:
+            break
+        overlap = any(is_overlapping(row, selected_tile) for selected_tile in selected_tiles)
+        if not overlap:
+            selected_tiles.append(row)
+    return pd.DataFrame(selected_tiles)
 
-import matplotlib.pyplot as plt
-import seaborn as sns
+def is_overlapping(tile1, tile2):
+    return not (tile1['x_end'] < tile2['x_start'] or tile1['x_start'] > tile2['x_end'] or
+                tile1['y_end'] < tile2['y_start'] or tile1['y_start'] > tile2['y_end'])
+
+
+imputed_predictions = pd.read_csv(
+    Path("results", "classifier", "informative_tiles", "exp", patient, "0", "experiment_1", "predictions",
+         f"{protein}_original_predictions.csv"))
+
+# get pre treatment predictions
+pre_tx_predictions = imputed_predictions[imputed_predictions["Treatment"] == "PRE"]
+pre_tx_predictions = pre_tx_predictions[
+    ["Treatment", "x_start", "x_end", "y_start", "y_end", "prediction_label"]]
+
+# get post treatment predictions
+post_tx_predictions = imputed_predictions[imputed_predictions["Treatment"] == "ON"]
+post_tx_predictions = post_tx_predictions[
+    ["Treatment", "x_start", "x_end", "y_start", "y_end", "prediction_label"]]
+
+# filter only wrong predictions
+wrong_pre_tx_predictions = pre_tx_predictions[
+    pre_tx_predictions["prediction_label"] != pre_tx_predictions["Treatment"]]
+wrong_post_tx_predictions = post_tx_predictions[
+    post_tx_predictions["prediction_label"] != post_tx_predictions["Treatment"]]
+
+# filter only correct predictions
+correct_pre_tx_predictions = pre_tx_predictions[
+    pre_tx_predictions["prediction_label"] == pre_tx_predictions["Treatment"]]
+correct_post_tx_predictions = post_tx_predictions[
+    post_tx_predictions["prediction_label"] == post_tx_predictions["Treatment"]]
+
+fig = sns.scatterplot(data=df, x="X_centroid", y="Y_centroid", s=10)
+# show only x 7000 to 8500, and y 4000 to 6000
+plt.xlim(7000, 8500)
+plt.ylim(4000, 6000)
+# rename X_centroid and Y_centroid to X and Y
+plt.xlabel("X")
+plt.ylabel("Y")
+count = 0
+
+correct_pre_tx_predictions = correct_pre_tx_predictions[correct_pre_tx_predictions["x_start"] < 8500]
+correct_pre_tx_predictions = correct_pre_tx_predictions[correct_pre_tx_predictions["x_start"] > 7000]
+correct_pre_tx_predictions = correct_pre_tx_predictions[correct_pre_tx_predictions["y_start"] < 6000]
+correct_pre_tx_predictions = correct_pre_tx_predictions[correct_pre_tx_predictions["y_start"] > 4000]
+
+# select 20 non overlapping tiles
+tiles = select_non_overlapping_tiles(correct_pre_tx_predictions, num_tiles=10)
+
+
+
+for i, row in tiles.iterrows():
+
+    x = row["x_start"]
+    y = row["y_start"]
 
-sns.scatterplot(data=df, x="X_centroid", y="Y_centroid", s=10)
+    width = row["x_end"] - row["x_start"]
+    height = row["y_end"] - row["y_start"]
+    fig.add_patch(plt.Rectangle((x, y), width, height, fill=False, edgecolor='green', lw=2))
 plt.show()