Update Pathology Pipelines for MetaTensor Changes (#905)

* Update tumor detection pipeline

Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

* Update MIL pipeline for MetaTensor changes

Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

Author: bhashemian

pathology/multiple_instance_learning/panda_mil_train_evaluate_pytorch_gpu.py

@@ -24,6 +24,7 @@
     RandGridPatchd,
     RandRotate90d,
     ScaleIntensityRanged,
+    SplitDimd,
     ToTensord,
 )
 from sklearn.metrics import cohen_kappa_score
@@ -235,9 +236,10 @@ def list_data_collate(batch: collections.abc.Sequence):
     """
 
     for i, item in enumerate(batch):
+        # print(f"{i} = {item['image'].shape=} >> {item['image'].keys=}")
         data = item[0]
         data["image"] = torch.stack([ix["image"] for ix in item], dim=0)
-        data["patch_location"] = torch.stack([ix["patch_location"] for ix in item], dim=0)
+        # data["patch_location"] = torch.stack([ix["patch_location"] for ix in item], dim=0)
         batch[i] = data
     return default_collate(batch)
 
@@ -289,6 +291,7 @@ def main_worker(gpu, args):
                 pad_mode=None,
                 constant_values=255,
             ),
+            SplitDimd(keys=["image"], dim=0, keepdim=False, list_output=True),
             RandFlipd(keys=["image"], spatial_axis=0, prob=0.5),
             RandFlipd(keys=["image"], spatial_axis=1, prob=0.5),
             RandRotate90d(keys=["image"], prob=0.5),
@@ -308,6 +311,7 @@ def main_worker(gpu, args):
                 pad_mode=None,
                 constant_values=255,
             ),
+            SplitDimd(keys=["image"], dim=0, keepdim=False, list_output=True),
             ScaleIntensityRanged(keys=["image"], a_min=np.float32(255), a_max=np.float32(0)),
             ToTensord(keys=["image", "label"]),
         ]

pathology/tumor_detection/ignite/camelyon_train_evaluate.py

@@ -115,7 +115,7 @@ def train(cfg):
     # Create MONAI dataset
     train_data_list = CSVDataset(
         cfg["train_file"],
-        col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+        col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
         kwargs_read_csv={"header": None},
         transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
     )
@@ -129,7 +129,7 @@ def train(cfg):
 
     valid_data_list = CSVDataset(
         cfg["valid_file"],
-        col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+        col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
         kwargs_read_csv={"header": None},
         transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
     )

pathology/tumor_detection/ignite/camelyon_train_evaluate_nvtx_profiling.py

@@ -117,7 +117,7 @@ def train(cfg):
     # Create MONAI dataset
     train_data_list = CSVDataset(
         cfg["train_file"],
-        col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+        col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
         kwargs_read_csv={"header": None},
         transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
     )
@@ -131,7 +131,7 @@ def train(cfg):
 
     valid_data_list = CSVDataset(
         cfg["valid_file"],
-        col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+        col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
         kwargs_read_csv={"header": None},
         transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
     )

pathology/tumor_detection/torch/camelyon_train_evaluate_pytorch_gpu.py

@@ -308,7 +308,7 @@ def main(cfg):
     # Create train dataset and dataloader
     train_data_list = CSVDataset(
         cfg["train_file"],
-        col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+        col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
         kwargs_read_csv={"header": None},
         transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
     )
@@ -327,7 +327,7 @@ def main(cfg):
     if not cfg["no_validate"]:
         valid_data_list = CSVDataset(
             cfg["valid_file"],
-            col_groups={"image": 0, "patch_location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
+            col_groups={"image": 0, "location": [2, 1], "label": [3, 6, 9, 4, 7, 10, 5, 8, 11]},
             kwargs_read_csv={"header": None},
             transform=Lambdad("image", lambda x: os.path.join(cfg["root"], "training/images", x + ".tif")),
         )
@@ -361,7 +361,7 @@ def main(cfg):
     # Deep Learning Model and Configurations
     # -------------------------------------------------------------------------
     # Initialize model
-    model = TorchVisionFCModel("resnet18", n_classes=1, use_conv=True, pretrained=cfg["pretrain"])
+    model = TorchVisionFCModel("resnet18", num_classes=1, use_conv=True, pretrained=cfg["pretrain"])
     model = model.to(device)
 
     # Loss function