Mercurial > repos > galaxyp > qupath_roi_splitter
changeset 6:6a8cf86fd3b7 draft default tip
planemo upload for repository hhttps://github.com/npinter/ROIsplitter commit 80eecd9912892296aad3231be75fcee3cc6a6b02
| author | galaxyp |
|---|---|
| date | Mon, 02 Sep 2024 19:07:23 +0000 |
| parents | 17c54a716a5b |
| children | |
| files | qupath_roi_splitter.py qupath_roi_splitter.xml |
| diffstat | 2 files changed, 50 insertions(+), 64 deletions(-) [+] |
line wrap: on
line diff
--- a/qupath_roi_splitter.py Tue Jul 30 12:59:11 2024 +0000 +++ b/qupath_roi_splitter.py Mon Sep 02 19:07:23 2024 +0000 @@ -14,29 +14,21 @@ return coords_with_index -def collect_roi_coords(input_roi, feature_index): - all_coords = [] - if len(input_roi["geometry"]["coordinates"]) == 1: - # Polygon w/o holes - all_coords.extend(collect_coords(input_roi["geometry"]["coordinates"][0], feature_index)) +def collect_roi_coords(input_roi): + coords = input_roi["geometry"]["coordinates"] + + def process_coords(coord_list): + if isinstance(coord_list[0], (int, float)): + return [coord_list] + elif all(isinstance(c, list) for c in coord_list): + return coord_list + else: + return [coord_list] + + if isinstance(coords[0][0], list): + return [process_coords(sub_coords) for sub_coords in coords] else: - coord_index = 0 - for sub_roi in input_roi["geometry"]["coordinates"]: - if len(sub_roi) == 2: - # Special case: LMD data - all_coords.extend(collect_coords([sub_roi], feature_index, coord_index)) - coord_index += 1 - else: - # Polygon with holes or MultiPolygon - if not isinstance(sub_roi[0][0], list): - all_coords.extend(collect_coords(sub_roi, feature_index, coord_index)) - coord_index += len(sub_roi) - else: - # MultiPolygon with holes - for sub_coord in sub_roi: - all_coords.extend(collect_coords(sub_coord, feature_index, coord_index)) - coord_index += len(sub_coord) - return all_coords + return [process_coords(coords)] def split_qupath_roi(in_roi): @@ -45,62 +37,56 @@ # HE dimensions dim_plt = [int(qupath_roi["dim"]["width"]), int(qupath_roi["dim"]["height"])] + tma_name = qupath_roi["name"] - tma_name = qupath_roi["name"] - cell_types = [ct.rsplit(" - ", 1)[-1] for ct in qupath_roi["featureNames"]] + if "featureNames" in qupath_roi: + cell_types = [ct.rsplit(" - ", 1)[-1] for ct in qupath_roi["featureNames"]] + else: + cell_types = ["all"] coords_by_cell_type = {ct: [] for ct in cell_types} - coords_by_cell_type['all'] = [] # For storing all coordinates if args.all is True + if "all" not in coords_by_cell_type: + coords_by_cell_type["all"] = [] - for feature_index, roi in enumerate(qupath_roi["features"]): - feature_coords = collect_roi_coords(roi, feature_index) + for roi in qupath_roi["features"]: + feature_coords = collect_roi_coords(roi) - if args.all: - coords_by_cell_type['all'].extend(feature_coords) + if args.all or "classification" not in roi["properties"]: + coords_by_cell_type["all"].append(feature_coords) elif "classification" in roi["properties"]: cell_type = roi["properties"]["classification"]["name"] if cell_type in cell_types: - coords_by_cell_type[cell_type].extend(feature_coords) + coords_by_cell_type[cell_type].append(feature_coords) - for cell_type, coords in coords_by_cell_type.items(): - if coords: - # Generate image (white background) - img = np.ones((dim_plt[1], dim_plt[0]), dtype="uint8") * 255 - - # Convert to numpy array and ensure integer coordinates - coords_arr = np.array(coords).astype(int) + for cell_type, coords_list in coords_by_cell_type.items(): + if coords_list: + img = np.ones((dim_plt[1], dim_plt[0], 3), dtype="uint8") * 255 - # Sort by feature_index first, then by coord_index - coords_arr = coords_arr[np.lexsort((coords_arr[:, 3], coords_arr[:, 2]))] + all_coords = [] + for feature in coords_list: + for polygon in feature: + # Multiple sub_polygons in LMD data + for sub_poly in polygon if isinstance(polygon[0][0], list) else [polygon]: + pts = np.array(sub_poly, dtype=np.float32).reshape(-1, 2) + pts = pts.astype(np.int32) - # Get filled pixel coordinates - if args.fill: - filled_coords = np.column_stack(np.where(img == 0)) - all_coords = np.unique(np.vstack((coords_arr[:, :2], filled_coords[:, ::-1])), axis=0) - else: - all_coords = coords_arr[:, :2] + # Get filled pixel coordinates + if args.fill: + temp_img = np.ones((dim_plt[1], dim_plt[0]), dtype="uint8") * 255 + cv2.fillPoly(temp_img, [pts], color=0) + filled_coords = np.column_stack(np.where(temp_img == 0)) + all_coords.extend(filled_coords[:, [1, 0]]) # Swap columns to get (x, y) + cv2.fillPoly(img, [pts], color=0) + else: + cv2.polylines(img, [pts], isClosed=True, color=(0, 0, 0), thickness=1) + all_coords.extend(pts) - # Save all coordinates to CSV + all_coords = np.array(all_coords) coords_df = pd.DataFrame(all_coords, columns=['x', 'y'], dtype=int) coords_df.to_csv("{}_{}.txt".format(tma_name, cell_type), sep='\t', index=False) # Generate image for visualization if --img is specified if args.img: - # Group coordinates by feature_index - features = {} - for x, y, feature_index, coord_index in coords_arr: - if feature_index not in features: - features[feature_index] = [] - features[feature_index].append((x, y)) - - # Draw each feature separately - for feature_coords in features.values(): - pts = np.array(feature_coords, dtype=np.int32) - if args.fill: - cv2.fillPoly(img, [pts], color=0) # Black fill - else: - cv2.polylines(img, [pts], isClosed=True, color=0, thickness=1) # Black outline - cv2.imwrite("{}_{}.png".format(tma_name, cell_type), img) @@ -108,7 +94,7 @@ parser = argparse.ArgumentParser(description="Split ROI coordinates of QuPath TMA annotation by cell type (classification)") parser.add_argument("--qupath_roi", default=False, help="Input QuPath annotation (GeoJSON file)") parser.add_argument("--fill", action="store_true", required=False, help="Fill pixels in ROIs (order of coordinates will be lost)") - parser.add_argument('--version', action='version', version='%(prog)s 0.3.1') + parser.add_argument('--version', action='version', version='%(prog)s 0.3.2') parser.add_argument("--all", action="store_true", required=False, help="Extracts all ROIs") parser.add_argument("--img", action="store_true", required=False, help="Generates image of ROIs") args = parser.parse_args()
--- a/qupath_roi_splitter.xml Tue Jul 30 12:59:11 2024 +0000 +++ b/qupath_roi_splitter.xml Mon Sep 02 19:07:23 2024 +0000 @@ -1,7 +1,7 @@ <tool id="qupath_roi_splitter" name="QuPath ROI Splitter" version="@VERSION@+galaxy@VERSION_SUFFIX@"> <description>Split ROI coordinates of QuPath TMA annotation by cell type (classification)</description> <macros> - <token name="@VERSION@">0.3.1</token> + <token name="@VERSION@">0.3.2</token> <token name="@VERSION_SUFFIX@">0</token> </macros> <requirements> @@ -67,7 +67,7 @@ <output_collection name="output_imgs" type="list" count="4"> <element name="E-5_Tumor.png"> <assert_contents> - <has_size value="459919"/> + <has_size value="1309478"/> </assert_contents> </element> </output_collection>