Refactor: minor README improvements + code syntax adjustments

README.md

@@ -8,9 +8,11 @@
 GPU-accelerated stain normalization tools for histopathological images. Compatible with PyTorch, TensorFlow, and Numpy.
 Normalization algorithms currently implemented:
 
-- Macenko [\[1\]](#reference) (ported from [numpy implementation](https://github.com/schaugf/HEnorm_python))
-- Reinhard [\[2\]](#reference)
-- Modified Reinhard [\[3\]](#reference)
+| Algorithm | numpy | torch | tensorflow |
+|-|-|-|-|
+| Macenko [\[1\]](#reference) | ✓ | ✓ | ✓ |
+| Reinhard [\[2\]](#reference) | ✓ | ✓ | ✓ |
+| Modified Reinhard [\[3\]](#reference) | ✓ | ✓ | ✓ |
 
 ## Installation
 
@@ -45,17 +47,9 @@ norm, H, E = normalizer.normalize(I=t_to_transform, stains=True)
 
 ![alt text](data/result.png)
 
-## Implemented algorithms
-
-| Algorithm | numpy | torch | tensorflow |
-|-|-|-|-|
-| Macenko | ✓ | ✓ | ✓ |
-| Reinhard | ✓ | ✓ | ✓ |
-| Modified Reinhard | ✓ | ✓ | ✓ |
-
 ## Backend comparison
 
-Results with 10 runs per size on a Intel(R) Core(TM) i5-8365U CPU @ 1.60GHz
+Macenko runtime results using different backends with 10 runs per image size on a Intel(R) Core(TM) i5-8365U CPU @ 1.60GHz.
 
 |   size | numpy avg. time   | torch avg. time   | tf avg. time     |
 |--------|-------------------|-------------------|------------------|
@@ -72,16 +66,15 @@ Results with 10 runs per size on a Intel(R) Core(TM) i5-8365U CPU @ 1.60GHz
 
 - [1] Macenko, Marc et al. "A method for normalizing histology slides for quantitative analysis." 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro. IEEE, 2009.
 - [2] Reinhard, Erik et al. "Color transfer between images." IEEE Computer Graphics and Applications. IEEE, 2001.
-- [3] Roy, Santanu et al. "Modified Reinhard Algorithm for Color Normalization of Colorectal Cancer Histopathology Images". 2021 29th European Signal Processing Conference (EUSIPCO), IEEE, 2021.
+- [3] Roy, Santanu et al. "Modified Reinhard Algorithm for Color Normalization of Colorectal Cancer Histopathology Images". 2021 29th European Signal Processing Conference (EUSIPCO). IEEE, 2021.
 
 ## Citing
 
 If you find this software useful for your research, please cite it as: 
 
 ```bibtex
 @software{barbano2022torchstain,
-  author       = {Carlo Alberto Barbano and
-                  André Pedersen},
+  author       = {Carlo Alberto Barbano and André Pedersen},
   title        = {EIDOSLAB/torchstain: v1.2.0-stable},
   month        = aug,
   year         = 2022,

compare.py

@@ -1,7 +1,6 @@
 import torch
 from torchvision import transforms
 import torchstain
-
 import cv2
 import matplotlib.pyplot as plt
 import time

example.py

@@ -5,7 +5,6 @@
 from torchvision import transforms
 import time
 
-
 size = 1024
 target = cv2.resize(cv2.cvtColor(cv2.imread("./data/target.png"), cv2.COLOR_BGR2RGB), (size, size))
 to_transform = cv2.resize(cv2.cvtColor(cv2.imread("./data/source.png"), cv2.COLOR_BGR2RGB), (size, size))

torchstain/numpy/normalizers/macenko.py

@@ -10,13 +10,13 @@ def __init__(self):
         super().__init__()
 
         self.HERef = np.array([[0.5626, 0.2159],
-                          [0.7201, 0.8012],
-                          [0.4062, 0.5581]])
+                               [0.7201, 0.8012],
+                               [0.4062, 0.5581]])
         self.maxCRef = np.array([1.9705, 1.0308])
 
     def __convert_rgb2od(self, I, Io=240, beta=0.15):
         # calculate optical density
-        OD = -np.log((I.astype(float)+1)/Io)
+        OD = -np.log((I.astype(float) + 1) / Io)
 
         # remove transparent pixels
         ODhat = OD[~np.any(OD < beta, axis=1)]
@@ -26,22 +26,22 @@ def __convert_rgb2od(self, I, Io=240, beta=0.15):
     def __find_HE(self, ODhat, eigvecs, alpha):
         #project on the plane spanned by the eigenvectors corresponding to the two
         # largest eigenvalues
-        That = ODhat.dot(eigvecs[:,1:3])
+        That = ODhat.dot(eigvecs[:, 1:3])
 
-        phi = np.arctan2(That[:,1],That[:,0])
+        phi = np.arctan2(That[:, 1], That[:, 0])
 
         minPhi = np.percentile(phi, alpha)
-        maxPhi = np.percentile(phi, 100-alpha)
+        maxPhi = np.percentile(phi, 100 - alpha)
 
-        vMin = eigvecs[:,1:3].dot(np.array([(np.cos(minPhi), np.sin(minPhi))]).T)
-        vMax = eigvecs[:,1:3].dot(np.array([(np.cos(maxPhi), np.sin(maxPhi))]).T)
+        vMin = eigvecs[:, 1:3].dot(np.array([(np.cos(minPhi), np.sin(minPhi))]).T)
+        vMax = eigvecs[:, 1:3].dot(np.array([(np.cos(maxPhi), np.sin(maxPhi))]).T)
 
         # a heuristic to make the vector corresponding to hematoxylin first and the
         # one corresponding to eosin second
         if vMin[0] > vMax[0]:
-            HE = np.array((vMin[:,0], vMax[:,0])).T
+            HE = np.array((vMin[:, 0], vMax[:, 0])).T
         else:
-            HE = np.array((vMax[:,0], vMin[:,0])).T
+            HE = np.array((vMax[:, 0], vMin[:, 0])).T
 
         return HE
 
@@ -55,7 +55,7 @@ def __find_concentration(self, OD, HE):
         return C
 
     def __compute_matrices(self, I, Io, alpha, beta):
-        I = I.reshape((-1,3))
+        I = I.reshape((-1, 3))
 
         OD, ODhat = self.__convert_rgb2od(I, Io=Io, beta=beta)
 
@@ -67,7 +67,7 @@ def __compute_matrices(self, I, Io, alpha, beta):
         C = self.__find_concentration(OD, HE)
 
         # normalize stain concentrations
-        maxC = np.array([np.percentile(C[0,:], 99), np.percentile(C[1,:],99)])
+        maxC = np.array([np.percentile(C[0, :], 99), np.percentile(C[1, :],99)])
 
         return HE, C, maxC
 
@@ -81,7 +81,7 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
         ''' Normalize staining appearence of H&E stained images
 
         Example use:
-            see test.py
+            see example.py
 
         Input:
             I: RGB input image
@@ -97,7 +97,7 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
             Macenko et al., ISBI 2009
         '''
         h, w, c = I.shape
-        I = I.reshape((-1,3))
+        I = I.reshape((-1, 3))
 
         HE, C, maxC = self.__compute_matrices(I, Io, alpha, beta)
 
@@ -114,11 +114,11 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
 
         if stains:
             # unmix hematoxylin and eosin
-            H = np.multiply(Io, np.exp(np.expand_dims(-self.HERef[:,0], axis=1).dot(np.expand_dims(C2[0,:], axis=0))))
+            H = np.multiply(Io, np.exp(np.expand_dims(-self.HERef[:, 0], axis=1).dot(np.expand_dims(C2[0, :], axis=0))))
             H[H > 255] = 255
             H = np.reshape(H.T, (h, w, c)).astype(np.uint8)
 
-            E = np.multiply(Io, np.exp(np.expand_dims(-self.HERef[:,1], axis=1).dot(np.expand_dims(C2[1,:], axis=0))))
+            E = np.multiply(Io, np.exp(np.expand_dims(-self.HERef[:, 1], axis=1).dot(np.expand_dims(C2[1, :], axis=0))))
             E[E > 255] = 255
             E = np.reshape(E.T, (h, w, c)).astype(np.uint8)
 

torchstain/tf/normalizers/macenko.py

@@ -4,7 +4,6 @@
 import numpy as np
 import tensorflow.keras.backend as K
 
-
 """
 Source code ported from: https://github.com/schaugf/HEnorm_python
 Original implementation: https://github.com/mitkovetta/staining-normalization
@@ -14,8 +13,8 @@ def __init__(self):
         super().__init__()
 
         self.HERef = tf.constant([[0.5626, 0.2159],
-                                   [0.7201, 0.8012],
-                                   [0.4062, 0.5581]])
+                                  [0.7201, 0.8012],
+                                  [0.4062, 0.5581]])
         self.maxCRef = tf.constant([1.9705, 1.0308])
 
     def __convert_rgb2od(self, I, Io, beta):
@@ -78,7 +77,7 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
         ''' Normalize staining appearence of H&E stained images
 
         Example use:
-            see test.py
+            see example.py
 
         Input:
             I: RGB input image: tensor of shape [C, H, W] and type uint8

torchstain/torch/normalizers/macenko.py

@@ -20,7 +20,7 @@ def __convert_rgb2od(self, I, Io, beta):
         I = I.permute(1, 2, 0)
 
         # calculate optical density
-        OD = -torch.log((I.reshape((-1, I.shape[-1])).float() + 1)/Io)
+        OD = -torch.log((I.reshape((-1, I.shape[-1])).float() + 1) / Io)
 
         # remove transparent pixels
         ODhat = OD[~torch.any(OD < beta, dim=1)]
@@ -79,7 +79,7 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
         ''' Normalize staining appearence of H&E stained images
 
         Example use:
-            see test.py
+            see example.py
 
         Input:
             I: RGB input image: tensor of shape [C, H, W] and type uint8

torchstain/torch/utils/rgb2lab.py

@@ -2,8 +2,8 @@
 
 # constant conversion matrices between color spaces: https://gist.github.com/bikz05/6fd21c812ef6ebac66e1
 _rgb2xyz = torch.tensor([[0.412453, 0.357580, 0.180423],
-                        [0.212671, 0.715160, 0.072169],
-                        [0.019334, 0.119193, 0.950227]])
+                         [0.212671, 0.715160, 0.072169],
+                         [0.019334, 0.119193, 0.950227]])
 
 _white = torch.tensor([0.95047, 1., 1.08883])