Implement a registry for cellfinder test data

README.md

@@ -49,3 +49,12 @@ See [our blog posts](https://brainglobe.info/blog/) for more information on the
 If you have encountered a bug whilst using cellfinder, please [open an issue on GitHub](https://github.com/brainglobe/cellfinder/issues).
 
 If you are interested in contributing to cellfinder (thank you!) - please head over to our [developer documentation](https://brainglobe.info/community/developers/index.html).
+
+## Citation
+If you find this package useful, and use it in your research, please cite the following paper:
+> Tyson, A. L., Rousseau, C. V., Niedworok, C. J., Keshavarzi, S., Tsitoura, C., Cossell, L., Strom, M. and Margrie, T. W. (2021) “A deep learning algorithm for 3D cell detection in whole mouse brain image datasets’ PLOS Computational Biology, 17(5), e1009074
+[https://doi.org/10.1371/journal.pcbi.1009074](https://doi.org/10.1371/journal.pcbi.1009074)
+
+**If you use this, or any other tools in the brainglobe suite, please
+ [let us know](https://brainglobe.info/contact.html), and
+ we'd be happy to promote your paper/talk etc.**

cellfinder/core/detect/filters/volume/structure_splitting.py

@@ -74,6 +74,7 @@ def ball_filter_imgs(
     good_tiles_mask = np.ones((1, 1, volume.shape[2]), dtype=np.bool_)
 
     plane_width, plane_height = volume.shape[:2]
+    current_z = ball_z_size // 2
 
     bf = BallFilter(
         plane_width,
@@ -86,9 +87,7 @@ def ball_filter_imgs(
         threshold_value=threshold_value,
         soma_centre_value=soma_centre_value,
     )
-    cell_detector = CellDetector(
-        plane_width, plane_height, start_z=ball_z_size // 2
-    )
+    cell_detector = CellDetector(plane_width, plane_height, start_z=current_z)
 
     # FIXME: hard coded type
     ball_filtered_volume = np.zeros(volume.shape, dtype=np.uint32)
@@ -98,7 +97,11 @@ def ball_filter_imgs(
         if bf.ready:
             bf.walk()
             middle_plane = bf.get_middle_plane()
-            ball_filtered_volume[:, :, z] = middle_plane[:]
+
+            # first valid middle plane is the current_z, not z
+            ball_filtered_volume[:, :, current_z] = middle_plane[:]
+            current_z += 1
+
             # DEBUG: TEST: transpose
             previous_plane = cell_detector.process(
                 middle_plane.copy(), previous_plane
@@ -134,7 +137,10 @@ def iterative_ball_filter(
         vol, cell_centres = ball_filter_imgs(
             vol, threshold_value, soma_centre_value
         )
-        vol -= 1
+
+        # vol is unsigned, so can't let zeros underflow to max value
+        vol[:, :, :] = np.where(vol != 0, vol - 1, 0)
+
         n_structures = len(cell_centres)
         ns.append(n_structures)
         centres.append(cell_centres)

pyproject.toml

@@ -44,6 +44,7 @@ cellfinder = "cellfinder.napari:napari.yaml"
 dev = [
     "black",
     "pre-commit",
+    "pooch >=1",
     "pyinstrument",
     "pytest-cov",
     "pytest-mock",

tests/core/conftest.py

@@ -1,7 +1,9 @@
 import os
+from pathlib import Path
 from typing import Tuple
 
 import numpy as np
+import pooch
 import pytest
 import torch.backends.mps
 from skimage.filters import gaussian
@@ -85,3 +87,63 @@ def synthetic_bright_spots() -> Tuple[np.ndarray, np.ndarray]:
     background_array = np.zeros_like(signal_array)
 
     return signal_array, background_array
+
+
+@pytest.fixture
+def test_data_registry():
+    """
+    Create a test data registry for BrainGlobe.
+
+    Returns:
+        pooch.Pooch: The test data registry object.
+
+    """
+    registry = pooch.create(
+        pooch.os_cache("brainglobe_test_data"),
+        base_url="https://gin.g-node.org/BrainGlobe/test-data/raw/master/",
+        env="BRAINGLOBE_TEST_DATA_DIR",
+    )
+    registry.load_registry(Path(__file__).parent / "registry.txt")
+    return registry
+
+
+@pytest.fixture
+def fetch_pooch_directory():
+    """
+    Fetches files from the Pooch registry that belong to a specific directory.
+
+    Parameters:
+        registry (pooch.Pooch): The Pooch registry object.
+        directory_name (str):
+            The remote relative path of the directory to fetch files from.
+        processor (callable, optional):
+            A function to process the fetched files. Defaults to None.
+        downloader (callable, optional):
+            A function to download the files. Defaults to None.
+        progressbar (bool, optional):
+            Whether to display a progress bar during the fetch.
+            Defaults to False.
+
+    Returns:
+        str: The local absolute path to the fetched directory.
+    """
+
+    def _fetch_pooch_registry(
+        registry: pooch.Pooch,
+        directory_name: str,
+        processor=None,
+        downloader=None,
+        progressbar=False,
+    ):
+        for name in registry.registry_files:
+            if name.startswith(f"{directory_name}/"):
+                registry.fetch(
+                    name,
+                    processor=processor,
+                    downloader=downloader,
+                    progressbar=progressbar,
+                )
+
+        return str(registry.abspath / directory_name)
+
+    return _fetch_pooch_registry

tests/core/registry.txt

@@ -0,0 +1,131 @@
+bright_brain/2d_filter/signal0001.tif f2cf4bb55ba9a57e99a01baf98830668cbbb1634619fc59c7d1e34dd1974b217
+bright_brain/2d_filter/signal0002.tif 7291fecc69430a746cde85248958b037c95d5f0c278900c6b4f2eb48447d8ec0
+bright_brain/2d_filter/signal0003.tif dec11106fcce6c19e3da8665614367bfca1499af1c805f7efa2fdefa31c55464
+bright_brain/2d_filter/signal0004.tif 2f03d5dbbfe9634847a82dbb2495777b127fcb7f385735fb8e2116ae0052a53f
+bright_brain/2d_filter/signal0005.tif 2fa6ed1d094560f3becb987c9eef847a1c95dc0f8b791c2ec7b5a7944207dd95
+bright_brain/3d_filter/plane_0002.tif ae8427a59696e5a35ae40d1b0998c4c4c478fbe46800c89f09b4556614ec0b3a
+bright_brain/3d_filter/plane_0003.tif d4a0e82b5e95b77c3b9d760591969ec99c5ac9ce7d835ea054c120a69a29fc56
+bright_brain/3d_filter/plane_0004.tif 312243944133fb8a58b86ea6ee762229582372c01b7c5c1761d56ef017856b15
+bright_brain/detected_cells.xml c13a4609d3b1f44c84d58e5f56610333769424bb07c87b23e8c5c62908045274
+bright_brain/info.txt 753533da65bf47c01791378c04972d4394061fbd8724d32b292a5d6bb1eedbf8
+bright_brain/peak_enhanced/signal0001.tif cf5008afd80ef665ae5f56798440c7e377936dbed83bcfa8989329f8eb9a1ff7
+bright_brain/peak_enhanced/signal0002.tif dc083752c40280d0fb05dbc384f25c5a1458f377f73cc5aeaaf86aa9b3fa50fc
+bright_brain/peak_enhanced/signal0003.tif a11708ce790b22d7abf2685d26a6a681fbf5acc9544902bf02da281e48d30d07
+bright_brain/peak_enhanced/signal0004.tif b1f12d130e8b429f4932d2d2ab2331944623daa15f4a2c3388dc341d1b26f881
+bright_brain/peak_enhanced/signal0005.tif 36a2c7136440ee0df0f76ff0385e869223cda94d1d1dfc156aa7bb5e59227397
+bright_brain/signal/signal0001.tif 1db2af2d72f62afe09ea99f682006bfee2814cf4bd83259f80428f0742d14f91
+bright_brain/signal/signal0002.tif 00d97024f67ca3dc455f7945f82520ceb991922c33f26318674a834a521b9960
+bright_brain/signal/signal0003.tif 31e2c7a93226d6e5077ca25a4da8c64065e4f8a330e15b7b771780f661b5c5dc
+bright_brain/signal/signal0004.tif a51660160329ecb63c0f4ddc28cf513c33a63de4d1a55b2212e49e57e1679cb0
+bright_brain/signal/signal0005.tif 16861fde570c2f8fc93447e1a43c911a0d126b4ed458d0a5ef1e52845ec23e0c
+bright_brain/tiles/signal0001.tif d76dba3ac06272bb29ac507fdf059ff6c7308ce04de77f621b09fc16753d71f0
+bright_brain/tiles/signal0002.tif d76dba3ac06272bb29ac507fdf059ff6c7308ce04de77f621b09fc16753d71f0
+bright_brain/tiles/signal0003.tif d76dba3ac06272bb29ac507fdf059ff6c7308ce04de77f621b09fc16753d71f0
+bright_brain/tiles/signal0004.tif d76dba3ac06272bb29ac507fdf059ff6c7308ce04de77f621b09fc16753d71f0
+bright_brain/tiles/signal0005.tif 609627eefd5f46e6034a2839663dcf5eabdddc5b921982f0f4feffaddb510694
+cellfinder-test-data.zip b0ef53b1530e4fa3128fcc0a752d0751909eab129d701f384fc0ea5f138c5914
+edge_cells_brain/2d_filter/signal0000.tif 4803df36346838928c2cc9afa5041cdc7f57fcac21eda55e1c3bd2f70761bd8d
+edge_cells_brain/2d_filter/signal0001.tif 39a008a23e3b91e10b0dd327a229550c85260cf57fd627a758877c668dd65cba
+edge_cells_brain/2d_filter/signal0002.tif 501a706b007872353b9a638b952c4499e4b7e41a421092d33b99f3ad37432eff
+edge_cells_brain/2d_filter/signal0003.tif 8c3f3ada6f6d7c108f778e8a0805a9d93dbd396d28b1e8a935a701b35e3797ef
+edge_cells_brain/2d_filter/signal0004.tif 81d203c47e2e1752620cc80d79047683a5a3947973b840f4afef56ec77f5e61d
+edge_cells_brain/2d_filter/signal0005.tif be0759a6601155d09f1698a67cb8b1c590a1deb51c9ece2d2fe52e9ff1eb2194
+edge_cells_brain/2d_filter/signal0006.tif 41acc77bbe3b4eb3379746939f659d6ec5b7b36f5cc9970427a9cc0cd59c2e25
+edge_cells_brain/2d_filter/signal0007.tif d8516e2a5c59b950536166525e9cefe2be81ed54527f92e9bf269431f471532a
+edge_cells_brain/2d_filter/signal0008.tif d59a4ac8275882fbb8b1947a07ae45716616302e35cdeb731529d1a8f521c298
+edge_cells_brain/2d_filter/signal0009.tif 2715c53774793b395cd1ee340eb3438617414d769c5f1e293fd5698588125960
+edge_cells_brain/2d_filter/signal0010.tif 525b30be7f86ae957c8e2edae1ecc958c869d21979c8452455356020df112523
+edge_cells_brain/2d_filter/signal0011.tif 83a808ac1be98329a3767213f3a4396721a33aa13cef252bfa8906b496173101
+edge_cells_brain/2d_filter/signal0012.tif 110042dfedbc8f0189ff1e9a04a5e01ae1cc4d5a0d593dc4884bdd4a7c1dbdaa
+edge_cells_brain/2d_filter/signal0013.tif c71930b8e6ed5d52b0c3699553faf75da728e343f766da174ee15057ff531e43
+edge_cells_brain/2d_filter/signal0014.tif d4b20626c940a35f5995be0b5ad36557634df3af947db10a24843ff325f94d1b
+edge_cells_brain/2d_filter/signal0015.tif 1f6ed840d97320421cb9cb9e48487126e428cca4135036642292f9b6572d98b8
+edge_cells_brain/2d_filter/signal0016.tif 6e975a52d7eb282ecffa7e0ed90d6c814347b6be2fde77842f80da97d1ede01c
+edge_cells_brain/2d_filter/signal0017.tif 4f3ef278746323db7cf585b4c40fead0925ce7a193af6a10ed84e330a9fd16a0
+edge_cells_brain/2d_filter/signal0018.tif f07233fe5134dff9f9935f3335b847aa1bd83e44ab379e45a48906fa4c1c7403
+edge_cells_brain/2d_filter/signal0019.tif c581dc2533e01f1e1c8d937c364f37c67af83ab3f85efb02eb9b1150a2e591dd
+edge_cells_brain/2d_filter/signal0020.tif 25224c7965908fa3a6229f15ff0f79db4e76e4e50d50b3813830f844a11fa529
+edge_cells_brain/3d_filter/plane_0002.tif 26a15013be4b4120a045da6209f384dc9551a11b2726c9f702cbf6ef1046f3d5
+edge_cells_brain/3d_filter/plane_0003.tif 0fd764081ea03d6a24a9ff167c6f5c648612914c4f9451a8d5be246704d9dcce
+edge_cells_brain/3d_filter/plane_0004.tif 6d09059ac4b4b67d02143223cb88b25925433a9bf9feea01826979cc697555a8
+edge_cells_brain/3d_filter/plane_0005.tif 3f71392ca5ee4c4f1164f8220ede6ffb1f6c15fa611e7ab258825929d9574601
+edge_cells_brain/3d_filter/plane_0006.tif 518320cd8b56a9f88adb647614672e8d02824125e89355facc49e6a98353bd8c
+edge_cells_brain/3d_filter/plane_0007.tif e0184c6791393856a6bb2c4955891992db2f66d6c82ceaf0b1ba63511363a195
+edge_cells_brain/3d_filter/plane_0008.tif 05eca0b171879e6e837325d299b98d6cac7c2134cd456b85dbc7ba64e51d5ee7
+edge_cells_brain/3d_filter/plane_0009.tif 25c398550efc467b82e89ecca8e8188082c0dbfca3e630d47b4e7a8f9cf77141
+edge_cells_brain/3d_filter/plane_0010.tif 9b02d5a99544245c5367e5f54a0daafd3f55be43bb66220903c6b6766144070f
+edge_cells_brain/3d_filter/plane_0011.tif 820b6064971a73ab5c346dda0710cf665ad87d9caf75a4d00440da56a0390e93
+edge_cells_brain/3d_filter/plane_0012.tif de1633ab4807bf61370d4ef7ff13d6cd73c8241030396013f977371a9291ade3
+edge_cells_brain/3d_filter/plane_0013.tif 819003634068868c17e6b9cca71a82607047e0e7aab01c4c61fa63df7355d7db
+edge_cells_brain/3d_filter/plane_0014.tif deae555248b8b03c6b7b1cd2b02213f13255645e83737decf9a920bae30d1508
+edge_cells_brain/3d_filter/plane_0015.tif 063e8370dfdf80202a052d370d86357b6f07e263139f1f9cb115968d96ca1e76
+edge_cells_brain/3d_filter/plane_0016.tif dead984c00ef4d14582b125797d2743ce774eec46242e855a58342820e47063e
+edge_cells_brain/3d_filter/plane_0017.tif d890497d0349c51379add2b188bded46ef528c56e016131da257fead119681ab
+edge_cells_brain/3d_filter/plane_0018.tif d75f0a2e767dfa46f0d2cc3f63c367391eac47b3076427bce688b39305d20387
+edge_cells_brain/3d_filter/plane_0019.tif e7ee64795b3f07c4cff4f8a2700ba0c1055604ce50039fe0590c900c6ece43b8
+edge_cells_brain/3d_filter/plane_0020.tif 7f94a93aa45877be81732c5be4f4d1d67bb1d4059bb6de833d60c146be621efc
+edge_cells_brain/detected_cells.xml 623df13b2ea1370713ab261d82691bb1d551a2fcc56645c87262e5276a27bee1
+edge_cells_brain/info.txt 779b34aae0ef50fad814e86409ee48839b2970c63eb5124548cfbc46126157fe
+edge_cells_brain/peak_enhanced/signal0000.tif 11dc924b3f30def59f6ca0b31a428a8cb454c163f4e6417d620659028ac544d9
+edge_cells_brain/peak_enhanced/signal0001.tif ac3605624094838c229f2d4ca2af9df9df17ba742b11d7ca684c42ce8e49557e
+edge_cells_brain/peak_enhanced/signal0002.tif a7759610b69fd0e9d023231488038482d8e810b9f4c096374ea267d2b7e2c5e7
+edge_cells_brain/peak_enhanced/signal0003.tif 25d712b49e7d97368c9f924876c23b132c0b0902f899ad5150c0b4bfe24d591b
+edge_cells_brain/peak_enhanced/signal0004.tif e8607d850649ccd68d8bcd0af12c9cfb98f191bb5c092b81ed9b6c87837b986e
+edge_cells_brain/peak_enhanced/signal0005.tif eff9cefc56c34f4307b9c94af90bcd6e41ef03d45855018eef9e40ab0bbdf116
+edge_cells_brain/peak_enhanced/signal0006.tif e5a058f759640b198069526772f972b9db0fb77d5502b5473a573d2f1a8fb746
+edge_cells_brain/peak_enhanced/signal0007.tif 44f7a1658c8c61f8e5b680ebb03d1f50779f0f22cf0066fb1b938764cd7353c2
+edge_cells_brain/peak_enhanced/signal0008.tif 5439cfeb6a6f17e6231bb532791a4d37c9fa00ea1c32b78184e8708938adf233
+edge_cells_brain/peak_enhanced/signal0009.tif 46c460f41e2ead3d694c9646ed4c21e6db57cae4fc3fd1fd4d411385877fdec7
+edge_cells_brain/peak_enhanced/signal0010.tif 0e74e27ef728f1c7b6db97172df8961917d9dba8c87cef045b7f2f523eec4da2
+edge_cells_brain/peak_enhanced/signal0011.tif 97495a61cfabec5df5a144305b394c306d56063307a911c7024d1b8e40089663
+edge_cells_brain/peak_enhanced/signal0012.tif 9809f14393cae19d7815c16370265e318ab13ea5188e075b73295a4968f071b5
+edge_cells_brain/peak_enhanced/signal0013.tif 136b1dd0b90be087fe87c1cf0caef02f136d54dc642aa91fc03a96944a113a0b
+edge_cells_brain/peak_enhanced/signal0014.tif a74e9bed032952209bdffdd9c351dff8f0bb46b4272999dd14822c254e824f2c
+edge_cells_brain/peak_enhanced/signal0015.tif eb675b9cabf241b2209473e62d84011349a997677a7e32b1d4a59b9f95229364
+edge_cells_brain/peak_enhanced/signal0016.tif 0df886c3fb7cc9a9dcd0c5803629ccd57125e5c9c574bca7445222cbb14d9be7
+edge_cells_brain/peak_enhanced/signal0017.tif 675125f17b0a4784aa5f56240abad9c155831f8fa3be0c6be13bd29cd1323a90
+edge_cells_brain/peak_enhanced/signal0018.tif be3df5edb2280eb69e05e94d7864d36635e929744667ae10e608f33b02f4363d
+edge_cells_brain/peak_enhanced/signal0019.tif 90ea965719aa7ad0f4a6480ab1e860d58ddc1a6bd6441365f6edf6f1e76edfab
+edge_cells_brain/peak_enhanced/signal0020.tif 7ba5bae6b3b8845cbbe0726da8d28138702e681167f6a99d050b063c4a271a23
+edge_cells_brain/signal/signal0000.tif 8214694de0e41dc6f9a456cc335cff60091129ebe700a1fe6f3a38ebbb02cc28
+edge_cells_brain/signal/signal0001.tif 6b7dec437f06c7a66f70d2c16dc8d5ec0f039a11b68a458e216574d281238d88
+edge_cells_brain/signal/signal0002.tif 4ebf6e8b96a970a64816210f9270fb19f5055cfe7fe6b2c3debd78485853c4b3
+edge_cells_brain/signal/signal0003.tif 920e5b95386d01766b94b357eb7e2906d2c6f1d27eee4e4765edb7455adcf91c
+edge_cells_brain/signal/signal0004.tif 90b84ea67c92d086f993588649dd6570a2baa46de6ceaf5aa64213673c922bbd
+edge_cells_brain/signal/signal0005.tif a6072c68f0636f4f7a34ada3ff3fad43105d2c7dbcf831aca6662a8b358a4d90
+edge_cells_brain/signal/signal0006.tif 87773e1f7bbbfad1324e5f83214e7339119b2e9b244a34d18f5f4211fd25978e
+edge_cells_brain/signal/signal0007.tif 2a91b8ba0128314753542a7890438b0805b039c3b74de80f8d967e4aa45f31e3
+edge_cells_brain/signal/signal0008.tif e2883d175bebc48f1cf7a82690f709d379715e471ccab740205040a203c3e88e
+edge_cells_brain/signal/signal0009.tif 854398ddc2de9f1f43256f935bb4bb2d3b36726ed615fedc045050adb618dda2
+edge_cells_brain/signal/signal0010.tif 8ccb209331784fe4aa8f43eb3174e68fee8abdc0d29acbb388ddf9b8236d8a31
+edge_cells_brain/signal/signal0011.tif 883722f089d5318881a2767ca4717857e52ccf0005e4535e09ed24de9ab9f73b
+edge_cells_brain/signal/signal0012.tif 127d14bb8e1f7a721d23f02f05d949345d78a1471e41270c8ad0cf012454fb95
+edge_cells_brain/signal/signal0013.tif b4cd368115d17048c81934dcb310e43663cc43ea14535516e75496e4622cee1d
+edge_cells_brain/signal/signal0014.tif 28c9d9dc3db609d48c69271f3755bcc5fbe3b870a8939b2c4685e7d8f3dfa8dd
+edge_cells_brain/signal/signal0015.tif 2388045546ed504281a190eddbe1a362d0f4f5803cc4cd3bac92a8e10ee5c7d6
+edge_cells_brain/signal/signal0016.tif 96f834737b8be786db9aca7edbdbae04d11c8f6c4deea51616a8567f39d1d03b
+edge_cells_brain/signal/signal0017.tif e79d979915c85598d1fc64e9d19c5a22f575a8c0ab7ae402f9da070027a19af9
+edge_cells_brain/signal/signal0018.tif af5d97c6365694fcff26bb04a37bab2c008b6d3ccacaaedf17ce4d003b289aec
+edge_cells_brain/signal/signal0019.tif bb798c254f2627e26ec53b110ad572e0494eddb06ce79d48de7267cb6e629f89
+edge_cells_brain/signal/signal0020.tif edcc27ab82449b3d11202a8866096e885bacab2dd63471fdb8e142293a9a99e8
+edge_cells_brain/tiles/signal0000.tif ea8bd86e6c8ee62e7de84a1f6792ba7d2174d3dffe9d1b578fbbde9c71dacc8c
+edge_cells_brain/tiles/signal0001.tif aa757c580996cd69264617ce3b2b7ac0a1173bac78fe5c5bdfbd7e61f633a764
+edge_cells_brain/tiles/signal0002.tif ea8bd86e6c8ee62e7de84a1f6792ba7d2174d3dffe9d1b578fbbde9c71dacc8c
+edge_cells_brain/tiles/signal0003.tif aa757c580996cd69264617ce3b2b7ac0a1173bac78fe5c5bdfbd7e61f633a764
+edge_cells_brain/tiles/signal0004.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0005.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0006.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0007.tif f382330f73c2625e84bc36a3faee169a074ed8e59db836fcbbd236061f3888b4
+edge_cells_brain/tiles/signal0008.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0009.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0010.tif d3560fff32b42f8d7baa32b2746535137675ade1bb3e726d8ef890e321daf8cc
+edge_cells_brain/tiles/signal0011.tif d3560fff32b42f8d7baa32b2746535137675ade1bb3e726d8ef890e321daf8cc
+edge_cells_brain/tiles/signal0012.tif d3560fff32b42f8d7baa32b2746535137675ade1bb3e726d8ef890e321daf8cc
+edge_cells_brain/tiles/signal0013.tif d3560fff32b42f8d7baa32b2746535137675ade1bb3e726d8ef890e321daf8cc
+edge_cells_brain/tiles/signal0014.tif d3560fff32b42f8d7baa32b2746535137675ade1bb3e726d8ef890e321daf8cc
+edge_cells_brain/tiles/signal0015.tif e0aaa58adcf0f13402494ce34112f7b937bf9421a1e0064856f57946e0267a02
+edge_cells_brain/tiles/signal0016.tif bc89f955fb25b3f536246912eaac757c7c3db4d9eddfdd5cb9b546800b782c2e
+edge_cells_brain/tiles/signal0017.tif b0a9789d16458fcfc92166c4363fe48859311247939587acb1d04d2885a36476
+edge_cells_brain/tiles/signal0018.tif 5307038034e5d4ca98c33557bc5ed9ef5ccb12672934e5387ff7e1d236f5cac7
+edge_cells_brain/tiles/signal0019.tif 40c4474b4ec25a8c68ac20f6649e7296af058e1bd854585ef7d71facb11a37e0
+edge_cells_brain/tiles/signal0020.tif b0a9789d16458fcfc92166c4363fe48859311247939587acb1d04d2885a36476

tests/core/test_integration/test_detection_structure_splitting.py

@@ -8,9 +8,13 @@
 
 import os
 
+import numpy as np
 import pytest
 from brainglobe_utils.IO.image.load import read_with_dask
 
+from cellfinder.core.detect.filters.volume.structure_splitting import (
+    split_cells,
+)
 from cellfinder.core.main import main
 
 data_dir = os.path.join(
@@ -46,3 +50,34 @@ def test_structure_splitting(signal_array, background_array):
         voxel_sizes,
         n_free_cpus=0,
     )
+
+
+def test_underflow_issue_435():
+    # two cells centered at (9, 10, 10), (19, 10, 10) with radius 5
+    p1 = np.array([9, 10, 10])
+    p2 = np.array([19, 10, 10])
+    radius = 5
+
+    bright_voxels = np.zeros((30, 20, 20), dtype=np.bool_)
+
+    pos = np.empty((30, 20, 20, 3))
+    pos[:, :, :, 0] = np.arange(30).reshape((-1, 1, 1))
+    pos[:, :, :, 1] = np.arange(20).reshape((1, -1, 1))
+    pos[:, :, :, 2] = np.arange(20).reshape((1, 1, -1))
+
+    dist1 = pos - p1.reshape((1, 1, 1, 3))
+    dist1 = np.sqrt(np.sum(np.square(dist1), axis=3))
+    inside1 = dist1 <= radius
+    dist2 = pos - p2.reshape((1, 1, 1, 3))
+    dist2 = np.sqrt(np.sum(np.square(dist2), axis=3))
+    inside2 = dist2 <= radius
+
+    bright_voxels[np.logical_or(inside1, inside2)] = True
+    bright_indices = np.argwhere(bright_voxels)
+
+    centers = split_cells(bright_indices)
+
+    # for some reason, same with pytorch, it's shifted by 1. Probably rounding
+    expected = {(10, 11, 11), (20, 11, 11)}
+    got = set(map(tuple, centers.tolist()))
+    assert expected == got