RuntimeError: applying transform ＜monai.transforms.croppad.dictionary.RandCropByPosNegLabeld object

项目场景：

最近在用UNETR做医学图像分割，在用官方的tutorial跑自己的数据集时出现了报错不能训练

---

问题描述

例如：在前面的transform设置，dataset和dataloader都能正常运行，在训练的时候下面这句代码

global_step, dice_val_best, global_step_best = train(
        global_step, train_loader, dice_val_best, global_step_best
    )

报错如下：

RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/transform.py", line 89, in apply_transform
    return _apply_transform(transform, data, unpack_items)
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/transform.py", line 53, in _apply_transform
    return transform(parameters)
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/croppad/dictionary.py", line 1171, in __call__
    self.randomize(label, fg_indices, bg_indices, image)
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/croppad/dictionary.py", line 1153, in randomize
    self.centers = generate_pos_neg_label_crop_centers(
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/utils.py", line 512, in generate_pos_neg_label_crop_centers
    centers.append(correct_crop_centers(center, spatial_size, label_spatial_shape, allow_smaller))
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/utils.py", line 439, in correct_crop_centers
    raise ValueError("The size of the proposed random crop ROI is larger than the image size.")
ValueError: The size of the proposed random crop ROI is larger than the image size.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 198, in _worker_loop
    data = fetcher.fetch(index)
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/opt/conda/lib/python3.8/site-packages/monai/data/dataset.py", line 97, in __getitem__
    return self._transform(index)
  File "/opt/conda/lib/python3.8/site-packages/monai/data/dataset.py", line 807, in _transform
    data = apply_transform(_transform, data)
  File "/opt/conda/lib/python3.8/site-packages/monai/transforms/transform.py", line 113, in apply_transform
    raise RuntimeError(f"applying transform {transform}") from e
RuntimeError: applying transform <monai.transforms.croppad.dictionary.RandCropByPosNegLabeld object at 0x7f3c221e9340>

• 不知道怎么解决把num_works改为了0依然报错。
• 查了这句报错：ValueError: The size of the proposed random crop ROI is larger than the image size参考.说是该错误是由于ROI裁剪过大而无法覆盖整个输入图像体积造成的。它与网络架构无关，但与数据加载器相关。于是找到transform设置中把spatial_size=(96, 96, 96)改成spatial_size=(48, 48, 48)

train_transforms = Compose(
    [
        LoadImaged(keys=["image", "label"]),
        AddChanneld(keys=["image", "label"]),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        Spacingd(
            keys=["image", "label"],
            pixdim=(1.5, 1.5, 2.0),
            mode=("bilinear", "nearest"),
        ),
        ScaleIntensityRanged(
            keys=["image"],
            a_min=-175,
            a_max=250,
            b_min=0.0,
            b_max=1.0,
            clip=True,
        ),
        CropForegroundd(keys=["image", "label"], source_key="image"),
        RandCropByPosNegLabeld(
            keys=["image", "label"],
            label_key="label",
#             spatial_size=(96, 96, 96),
            spatial_size=(48, 48, 48),
            pos=1,
            neg=1,
            num_samples=4,
            image_key="image",
            image_threshold=0,
        ),
        RandFlipd(
            keys=["image", "label"],
            spatial_axis=[0],
            prob=0.10,
        ),
        RandFlipd(
            keys=["image", "label"],
            spatial_axis=[1],
            prob=0.10,
        ),
        RandFlipd(
            keys=["image", "label"],
            spatial_axis=[2],
            prob=0.10,
        ),
        RandRotate90d(
            keys=["image", "label"],
            prob=0.10,
            max_k=3,
        ),
        RandShiftIntensityd(
            keys=["image"],
            offsets=0.10,
            prob=0.50,
        ),
        ToTensord(keys=["image", "label"]),
    ]
)

但是训练模型时又报了新的错误，报错信息如下：

RuntimeError: The size of tensor a (27) must match the size of tensor b (216) at non-singleton dimension 1

---

原因分析：

ROI裁剪尺寸大于图片本身了，

---

解决方案：

在调用UNETR模型时，参数img_size=(96, 96, 96)改为img_size=(32, 32, 32)