Feat/ikeda042 cell pos metadata

README.md

@@ -233,7 +233,7 @@ images or raw intensity data through the analysis pipeline.
 | Uploaded ND2 files | `backend/app/nd2files/` |
 | Cell databases | `backend/app/databases/<nd2_stem>.db` |
 | Extracted contour previews | `backend/app/extracted_data/<nd2_stem>/<frame>.png` |
-| Cell table | SQLite table `cells`, including `cell_id`, `manual_label`, `perimeter`, `area`, `img_ph`, `img_fluo1`, `img_fluo2`, `contour`, center coordinates, objective, and pixel size. |
+| Cell table | SQLite table `cells`, including `cell_id`, `manual_label`, `perimeter`, `area`, `img_ph`, `img_fluo1`, `img_fluo2`, `contour`, crop-local center coordinates, original ND2 frame position coordinates, objective, and pixel size. |
 | Bulk exports | JSON/CSV/PNG responses from Bulk Engine endpoints or browser downloads. |
 | Frontend production build | `frontend/dist/`; the backend serves this folder when it exists. |
 

backend/app/cellextraction/README.md

@@ -98,8 +98,8 @@ Failure response:
 - Databases: SQLite files written to `backend/app/databases/`.
   - Default name: `<nd2_stem>.db` with dots in the filename replaced by `p`.
   - Table: `cells` (stores `cell_id`, `perimeter`, `area`, `img_ph`, `img_fluo1`,
-    `img_fluo2`, `contour`, `center_x`, `center_y`, and labels).
-  - With `auto_annotation=true`, `manual_label` is initialized to `1` or `N/A`.
+    `img_fluo2`, `contour`, crop-local `center_x`/`center_y`, original ND2
+    frame `position_x`/`position_y`, and labels).
 - Contour overlays: PNGs written to `backend/app/extracted_data/<nd2_stem>/`.
   - Files are named by frame index, e.g. `0.png`, `1.png`, `2.png`.
 

backend/app/cellextraction/crud.py

@@ -3,6 +3,7 @@
 import pickle
 import random
 import re
+import json
 import logging
 from functools import lru_cache
 from dataclasses import dataclass
@@ -35,6 +36,7 @@
 DATABASES_DIR: Path = APP_DIR / "databases"
 EXTRACTED_DATA_DIR: Path = APP_DIR / "extracted_data"
 TEMPDATA_DIR: Path = APP_DIR / "tempdata"
+CELL_POSITIONS_FILENAME: str = "cell_positions.json"
 AUTOANNOTATION_MODEL_PATH: Path = (
     APP_DIR.parent / "autoannotation" / "artifacts" / "autoannotator.pkl"
 )
@@ -302,6 +304,8 @@ class Cell(Base):
     contour = Column(BLOB)
     center_x = Column(FLOAT)
     center_y = Column(FLOAT)
+    position_x = Column(FLOAT, nullable=True)
+    position_y = Column(FLOAT, nullable=True)
     user_id = Column(String, nullable=True)
     objective_magnification = Column(String, nullable=True)
     pixel_size_um = Column(FLOAT, nullable=True)
@@ -681,6 +685,7 @@ def _ensure_dir(path: str) -> None:
                 _ensure_dir(path)
         loop_num = num_tiff // set_num if mode != "single_layer" else num_tiff
         for k in range(loop_num):
+            frame_dir = f"{temp_dir}/frames/tiff_{k}"
             image_ph_raw = SyncChores.load_image_unchanged(f"{temp_dir}/PH/{k}.tif")
             if image_ph_raw is None:
                 continue
@@ -740,9 +745,10 @@ def _ensure_dir(path: str) -> None:
             cv2.drawContours(image_ph_copy, contours, -1, (0, 255, 0), 3)
             cv2.imwrite(f"{contour_dir}/{k}.png", image_ph_copy)
             n = 0
+            saved_positions: dict[str, dict[str, float]] = {}
             if mode in ("triple_layer", "quad_layer"):
-                for ph, fluo1, fluo2 in zip(
-                    cropped_images_ph, cropped_images_fluo_1, cropped_images_fluo_2
+                for candidate_index, (ph, fluo1, fluo2) in enumerate(
+                    zip(cropped_images_ph, cropped_images_fluo_1, cropped_images_fluo_2)
                 ):
                     if (
                         len(ph) == output_size[0]
@@ -759,21 +765,41 @@ def _ensure_dir(path: str) -> None:
                         cv2.imwrite(
                             f"{temp_dir}/frames/tiff_{k}/Cells/fluo2/{n}.png", fluo2
                         )
+                        cx, cy = contour_centers[candidate_index]
+                        saved_positions[str(n)] = {
+                            "position_x": float(cx),
+                            "position_y": float(cy),
+                        }
                         n += 1
 
             elif mode == "single_layer":
-                for ph in cropped_images_ph:
+                for candidate_index, ph in enumerate(cropped_images_ph):
                     if len(ph) == output_size[0] and len(ph[0]) == output_size[1]:
                         cv2.imwrite(f"{temp_dir}/frames/tiff_{k}/Cells/ph/{n}.png", ph)
+                        cx, cy = contour_centers[candidate_index]
+                        saved_positions[str(n)] = {
+                            "position_x": float(cx),
+                            "position_y": float(cy),
+                        }
                         n += 1
             elif mode == "dual_layer":
-                for ph, fluo1 in zip(cropped_images_ph, cropped_images_fluo_1):
+                for candidate_index, (ph, fluo1) in enumerate(
+                    zip(cropped_images_ph, cropped_images_fluo_1)
+                ):
                     if len(ph) == output_size[0] and len(ph[0]) == output_size[1]:
                         cv2.imwrite(f"{temp_dir}/frames/tiff_{k}/Cells/ph/{n}.png", ph)
                         cv2.imwrite(
                             f"{temp_dir}/frames/tiff_{k}/Cells/fluo1/{n}.png", fluo1
                         )
+                        cx, cy = contour_centers[candidate_index]
+                        saved_positions[str(n)] = {
+                            "position_x": float(cx),
+                            "position_y": float(cy),
+                        }
                         n += 1
+            positions_path = Path(frame_dir) / "Cells" / CELL_POSITIONS_FILENAME
+            with positions_path.open("w", encoding="utf-8") as handle:
+                json.dump(saved_positions, handle, ensure_ascii=True, indent=2)
         return num_tiff
 
 
@@ -848,6 +874,33 @@ def process_image(
         contour = contours[0] if contours else None
         return contour, img_ph_gray, img_fluo1_gray, img_fluo2_gray
 
+    def get_cell_original_position(
+        self,
+        i: int,
+        j: int,
+    ) -> tuple[float | None, float | None]:
+        positions_path = (
+            Path(self.temp_dir)
+            / "frames"
+            / f"tiff_{i}"
+            / "Cells"
+            / CELL_POSITIONS_FILENAME
+        )
+        try:
+            with positions_path.open("r", encoding="utf-8") as handle:
+                positions = json.load(handle)
+        except (FileNotFoundError, json.JSONDecodeError, TypeError):
+            return None, None
+
+        entry = positions.get(str(j)) if isinstance(positions, dict) else None
+        if not isinstance(entry, dict):
+            return None, None
+
+        try:
+            return float(entry["position_x"]), float(entry["position_y"])
+        except (KeyError, TypeError, ValueError):
+            return None, None
+
     def process_cell(
         self,
         i: int,
@@ -876,6 +929,7 @@ def process_cell(
             or abs(center_y - img_ph.shape[0] // 2) >= 3
         ):
             return None
+        position_x, position_y = self.get_cell_original_position(i, j)
 
         img_ph_data = cv2.imencode(".png", img_ph_gray)[1].tobytes()
         img_fluo1_data = img_fluo2_data = None
@@ -914,6 +968,8 @@ def process_cell(
             contour=contour_blob,
             center_x=center_x,
             center_y=center_y,
+            position_x=position_x,
+            position_y=position_y,
             user_id=user_id,
             objective_magnification=self.objective_magnification,
             pixel_size_um=self.pixel_size_um,