ScaleIntensityRange, ScaleIntensityRanged, ScaleIntensityRangePercent…

…iles, ScaleIntensityRangePercentilesd (#2943)

ScaleIntensityRange, ScaleIntensityRanged, ScaleIntensityRangePercentiles, ScaleIntensityRangePercentilesd

monai/transforms/__init__.py

@@ -524,4 +524,4 @@
     weighted_patch_samples,
     zero_margins,
 )
-from .utils_pytorch_numpy_unification import in1d, moveaxis, where
+from .utils_pytorch_numpy_unification import clip, in1d, moveaxis, percentile, where

monai/transforms/intensity/array.py

@@ -28,7 +28,7 @@
 from monai.networks.layers import GaussianFilter, HilbertTransform, SavitzkyGolayFilter
 from monai.transforms.transform import RandomizableTransform, Transform
 from monai.transforms.utils import Fourier, equalize_hist, is_positive, rescale_array
-from monai.transforms.utils_pytorch_numpy_unification import where
+from monai.transforms.utils_pytorch_numpy_unification import clip, percentile, where
 from monai.utils import (
     PT_BEFORE_1_7,
     InvalidPyTorchVersionError,
@@ -688,14 +688,16 @@ class ScaleIntensityRange(Transform):
         clip: whether to perform clip after scaling.
     """
 
+    backend = [TransformBackends.TORCH, TransformBackends.NUMPY]
+
     def __init__(self, a_min: float, a_max: float, b_min: float, b_max: float, clip: bool = False) -> None:
         self.a_min = a_min
         self.a_max = a_max
         self.b_min = b_min
         self.b_max = b_max
         self.clip = clip
 
-    def __call__(self, img: np.ndarray):
+    def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
         """
         Apply the transform to `img`.
         """
@@ -706,7 +708,7 @@ def __call__(self, img: np.ndarray):
         img = (img - self.a_min) / (self.a_max - self.a_min)
         img = img * (self.b_max - self.b_min) + self.b_min
         if self.clip:
-            img = np.asarray(np.clip(img, self.b_min, self.b_max))
+            img = clip(img, self.b_min, self.b_max)
         return img
 
 
@@ -835,6 +837,8 @@ class ScaleIntensityRangePercentiles(Transform):
         relative: whether to scale to the corresponding percentiles of [b_min, b_max].
     """
 
+    backend = ScaleIntensityRange.backend
+
     def __init__(
         self, lower: float, upper: float, b_min: float, b_max: float, clip: bool = False, relative: bool = False
     ) -> None:
@@ -849,12 +853,12 @@ def __init__(
         self.clip = clip
         self.relative = relative
 
-    def __call__(self, img: np.ndarray):
+    def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
         """
         Apply the transform to `img`.
         """
-        a_min = np.percentile(img, self.lower)
-        a_max = np.percentile(img, self.upper)
+        a_min: float = percentile(img, self.lower)  # type: ignore
+        a_max: float = percentile(img, self.upper)  # type: ignore
         b_min = self.b_min
         b_max = self.b_max
 
@@ -866,7 +870,7 @@ def __call__(self, img: np.ndarray):
         img = scalar(img)
 
         if self.clip:
-            img = np.asarray(np.clip(img, self.b_min, self.b_max))
+            img = clip(img, self.b_min, self.b_max)
 
         return img
 

monai/transforms/intensity/dictionary.py

@@ -699,6 +699,8 @@ class ScaleIntensityRanged(MapTransform):
         allow_missing_keys: don't raise exception if key is missing.
     """
 
+    backend = ScaleIntensityRange.backend
+
     def __init__(
         self,
         keys: KeysCollection,
@@ -712,7 +714,7 @@ def __init__(
         super().__init__(keys, allow_missing_keys)
         self.scaler = ScaleIntensityRange(a_min, a_max, b_min, b_max, clip)
 
-    def __call__(self, data: Mapping[Hashable, np.ndarray]) -> Dict[Hashable, np.ndarray]:
+    def __call__(self, data: Mapping[Hashable, NdarrayOrTensor]) -> Dict[Hashable, NdarrayOrTensor]:
         d = dict(data)
         for key in self.key_iterator(d):
             d[key] = self.scaler(d[key])
@@ -816,6 +818,8 @@ class ScaleIntensityRangePercentilesd(MapTransform):
         allow_missing_keys: don't raise exception if key is missing.
     """
 
+    backend = ScaleIntensityRangePercentiles.backend
+
     def __init__(
         self,
         keys: KeysCollection,
@@ -830,7 +834,7 @@ def __init__(
         super().__init__(keys, allow_missing_keys)
         self.scaler = ScaleIntensityRangePercentiles(lower, upper, b_min, b_max, clip, relative)
 
-    def __call__(self, data: Mapping[Hashable, np.ndarray]) -> Dict[Hashable, np.ndarray]:
+    def __call__(self, data: Mapping[Hashable, NdarrayOrTensor]) -> Dict[Hashable, NdarrayOrTensor]:
         d = dict(data)
         for key in self.key_iterator(d):
             d[key] = self.scaler(d[key])

monai/transforms/utils_pytorch_numpy_unification.py

@@ -9,6 +9,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from typing import Union
+
 import numpy as np
 import torch
 
@@ -17,6 +19,8 @@
 __all__ = [
     "moveaxis",
     "in1d",
+    "clip",
+    "percentile",
     "where",
 ]
 
@@ -53,6 +57,56 @@ def in1d(x, y):
     return (x[..., None] == torch.tensor(y, device=x.device)).any(-1).view(-1)
 
 
+def clip(a: NdarrayOrTensor, a_min, a_max) -> NdarrayOrTensor:
+    """`np.clip` with equivalent implementation for torch."""
+    result: NdarrayOrTensor
+    if isinstance(a, np.ndarray):
+        result = np.clip(a, a_min, a_max)
+    else:
+        result = torch.clip(a, a_min, a_max)
+    return result
+
+
+def percentile(x: NdarrayOrTensor, q) -> Union[NdarrayOrTensor, float, int]:
+    """`np.percentile` with equivalent implementation for torch.
+
+    Pytorch uses `quantile`, but this functionality is only available from v1.7.
+    For earlier methods, we calculate it ourselves. This doesn't do interpolation,
+    so is the equivalent of ``numpy.percentile(..., interpolation="nearest")``.
+
+    Args:
+        x: input data
+        q: percentile to compute (should in range 0 <= q <= 100)
+
+    Returns:
+        Resulting value (scalar)
+    """
+    if np.isscalar(q):
+        if not 0 <= q <= 100:
+            raise ValueError
+    else:
+        if any(q < 0) or any(q > 100):
+            raise ValueError
+    result: Union[NdarrayOrTensor, float, int]
+    if isinstance(x, np.ndarray):
+        result = np.percentile(x, q)
+    else:
+        q = torch.tensor(q, device=x.device)
+        if hasattr(torch, "quantile"):
+            result = torch.quantile(x, q / 100.0)
+        else:
+            # Note that ``kthvalue()`` works one-based, i.e., the first sorted value
+            # corresponds to k=1, not k=0. Thus, we need the `1 +`.
+            k = 1 + (0.01 * q * (x.numel() - 1)).round().int()
+            if k.numel() > 1:
+                r = [x.view(-1).kthvalue(int(_k)).values.item() for _k in k]
+                result = torch.tensor(r, device=x.device)
+            else:
+                result = x.view(-1).kthvalue(int(k)).values.item()
+
+    return result
+
+
 def where(condition: NdarrayOrTensor, x, y) -> NdarrayOrTensor:
     """
     Note that `torch.where` may convert y.dtype to x.dtype.

tests/test_scale_intensity_range.py

@@ -11,19 +11,18 @@
 
 import unittest
 
-import numpy as np
-
 from monai.transforms import ScaleIntensityRange
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class IntensityScaleIntensityRange(NumpyImageTestCase2D):
     def test_image_scale_intensity_range(self):
         scaler = ScaleIntensityRange(a_min=20, a_max=108, b_min=50, b_max=80)
-        scaled = scaler(self.imt)
-        expected = (self.imt - 20) / 88
-        expected = expected * 30 + 50
-        self.assertTrue(np.allclose(scaled, expected))
+        for p in TEST_NDARRAYS:
+            scaled = scaler(p(self.imt))
+            expected = (self.imt - 20) / 88
+            expected = expected * 30 + 50
+            assert_allclose(scaled, expected)
 
 
 if __name__ == "__main__":

tests/test_scale_intensity_range_percentiles.py

@@ -14,7 +14,7 @@
 import numpy as np
 
 from monai.transforms.intensity.array import ScaleIntensityRangePercentiles
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestScaleIntensityRangePercentiles(NumpyImageTestCase2D):
@@ -30,7 +30,9 @@ def test_scaling(self):
         expected = (img - a_min) / (a_max - a_min)
         expected = (expected * (b_max - b_min)) + b_min
         scaler = ScaleIntensityRangePercentiles(lower=lower, upper=upper, b_min=b_min, b_max=b_max)
-        self.assertTrue(np.allclose(expected, scaler(img)))
+        for p in TEST_NDARRAYS:
+            result = scaler(p(img))
+            assert_allclose(expected, result)
 
     def test_relative_scaling(self):
         img = self.imt
@@ -47,7 +49,9 @@ def test_relative_scaling(self):
         expected_img = (img - expected_a_min) / (expected_a_max - expected_a_min)
         expected_img = (expected_img * (expected_b_max - expected_b_min)) + expected_b_min
 
-        self.assertTrue(np.allclose(expected_img, scaler(img)))
+        for p in TEST_NDARRAYS:
+            result = scaler(p(img))
+            assert_allclose(expected_img, result)
 
     def test_invalid_instantiation(self):
         self.assertRaises(ValueError, ScaleIntensityRangePercentiles, lower=-10, upper=99, b_min=0, b_max=255)

tests/test_scale_intensity_ranged.py

@@ -11,20 +11,19 @@
 
 import unittest
 
-import numpy as np
-
 from monai.transforms import ScaleIntensityRanged
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class IntensityScaleIntensityRanged(NumpyImageTestCase2D):
     def test_image_scale_intensity_ranged(self):
         key = "img"
         scaler = ScaleIntensityRanged(keys=key, a_min=20, a_max=108, b_min=50, b_max=80)
-        scaled = scaler({key: self.imt})
-        expected = (self.imt - 20) / 88
-        expected = expected * 30 + 50
-        self.assertTrue(np.allclose(scaled[key], expected))
+        for p in TEST_NDARRAYS:
+            scaled = scaler({key: p(self.imt)})
+            expected = (self.imt - 20) / 88
+            expected = expected * 30 + 50
+            assert_allclose(scaled[key], expected)
 
 
 if __name__ == "__main__":

tests/test_utils_pytorch_numpy_unification.py

@@ -0,0 +1,46 @@
+# Copyright 2020 - 2021 MONAI Consortium
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#     http://www.apache.org/licenses/LICENSE-2.0
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+import numpy as np
+import torch
+
+from monai.transforms.utils_pytorch_numpy_unification import percentile
+from tests.utils import TEST_NDARRAYS, assert_allclose, set_determinism
+
+
+class TestPytorchNumpyUnification(unittest.TestCase):
+    def setUp(self) -> None:
+        set_determinism(0)
+
+    def test_percentile(self):
+        for size in (1, 100):
+            q = np.random.randint(0, 100, size=size)
+            results = []
+            for p in TEST_NDARRAYS:
+                arr = p(np.arange(100 * 101).reshape(1, 100, 101).astype(np.float32))
+                results.append(percentile(arr, q))
+                # pre torch 1.7, no `quantile`. Our own method doesn't interpolate,
+                # so we can only be accurate to 0.5
+                atol = 0.5 if not hasattr(torch, "quantile") else 1e-4
+                assert_allclose(results[0], results[-1], atol=atol)
+
+    def test_fails(self):
+        for p in TEST_NDARRAYS:
+            for q in (-1, 101):
+                arr = p(np.arange(100 * 101).reshape(1, 100, 101).astype(np.float32))
+                with self.assertRaises(ValueError):
+                    percentile(arr, q)
+
+
+if __name__ == "__main__":
+    unittest.main()