more conservative SV filtering at z borders

Author: SarahMuth

run_sbatch_revision.sbatch

@@ -1,15 +1,16 @@
 #! /bin/bash
 #SBATCH -c 4 #4 #8
-#SBATCH --mem 120G #120G #32G #64G #256G
+#SBATCH --mem 256G #120G #32G #64G #256G
 #SBATCH -p grete:shared #grete:shared #grete-h100:shared
-#SBATCH -t 4:00:00 #6:00:00 #48:00:00 
+#SBATCH -t 3:00:00 #6:00:00 #48:00:00 
 #SBATCH -G A100:1 #V100:1 #2 #A100:1  #gtx1080:2 #v100:1 #H100:1
 #SBATCH --output=/user/muth9/u12095/synapse-net/slurm_revision/slurm-%j.out 
-#SBATCH -A nim00007 #SBATCH --constraint 80gb 
+#SBATCH -A nim00007 
+#SBATCH --constraint 80gb 
 
 source ~/.bashrc
 conda activate synapse-net
 python /user/muth9/u12095/synapse-net/scripts/cooper/revision/updated_data_analysis/run_data_analysis.py \
- -i /mnt/lustre-emmy-hdd/projects/nim00007/data/synaptic-reconstruction/cooper/20241102_TOMO_DATA_Imig2014/exported/SNAP25/ \
- -o /mnt/lustre-emmy-hdd/projects/nim00007/data/synaptic-reconstruction/cooper/20241102_TOMO_DATA_Imig2014/afterRevision_analysis/boundaryT0_9_constantins_presynapticFiltering --store \
- -s ./analysis_results/man_subset
+ -i /mnt/lustre-emmy-hdd/projects/nim00007/data/synaptic-reconstruction/cooper/20241102_TOMO_DATA_Imig2014/exported/SNAP25 \
+ -o /mnt/lustre-emmy-hdd/projects/nim00007/data/synaptic-reconstruction/cooper/20241102_TOMO_DATA_Imig2014/afterRevision_analysis/boundaryT0_9_constantins_presynapticFiltering/weaker_SVfilter/man_subset --store \
+ -s ./analysis_results/weaker_SVfilter/man_subset

scripts/cooper/revision/updated_data_analysis/analysis_segmentations.py

@@ -71,7 +71,7 @@ def SV_pred(raw: np.ndarray, SV_model: str, output_path: str = None, store: bool
 
     if not use_existing_seg:
         #Excluding boundary SV, because they would also not be used in the manual annotation
-        seg, pred = segment_vesicles(input_volume=raw, model_path=SV_model, exclude_boundary=True, verbose=False, return_predictions=True)
+        seg, pred = segment_vesicles(input_volume=raw, model_path=SV_model, exclude_boundary_vesicles=True, verbose=False, return_predictions=True)
 
         if store and output_path:
             with h5py.File(output_path, "a") as f:

synapse_net/inference/vesicles.py

@@ -7,7 +7,7 @@
 import torch
 
 from synapse_net.inference.util import apply_size_filter, get_prediction, _Scaler
-from synapse_net.inference.postprocessing.vesicles import filter_border_objects
+from synapse_net.inference.postprocessing.vesicles import filter_border_objects, filter_border_vesicles
 
 
 def distance_based_vesicle_segmentation(
@@ -132,6 +132,7 @@ def segment_vesicles(
     return_predictions: bool = False,
     scale: Optional[List[float]] = None,
     exclude_boundary: bool = False,
+    exclude_boundary_vesicles: bool = False,
     mask: Optional[np.ndarray] = None,
 ) -> Union[np.ndarray, Tuple[np.ndarray, np.ndarray]]:
     """Segment vesicles in an input volume or image.
@@ -182,6 +183,25 @@ def segment_vesicles(
 
     if exclude_boundary:
         seg = filter_border_objects(seg)
+    if exclude_boundary_vesicles:
+        seg_ids = filter_border_vesicles(seg)
+        # Step 1: Zero out everything not in seg_ids
+        seg[~np.isin(seg, seg_ids)] = 0
+
+        # Step 2: Relabel remaining IDs to be consecutive starting from 1
+        unique_ids = np.unique(seg)
+        unique_ids = unique_ids[unique_ids != 0]  # Exclude background (0)
+
+        label_map = {old_label: new_label for new_label, old_label in enumerate(unique_ids, start=1)}
+
+        # Apply relabeling using a temp array (to avoid large ints in-place)
+        new_seg = np.zeros_like(seg, dtype=np.int32)
+        for old_label, new_label in label_map.items():
+            new_seg[seg == old_label] = new_label
+
+        # Final step: replace original seg with relabelled and casted version
+        seg = new_seg
+
     seg = scaler.rescale_output(seg, is_segmentation=True)
 
     if return_predictions: