Merge pull request #64 from andreped/fix/multitarget

Fixed bug in multitarget macenko during import

Author: carloalbertobarbano

README.md

@@ -1,7 +1,7 @@
 # torchstain
 
 [![License](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
-[![tests](https://github.com/EIDOSLAB/torchstain/workflows/tests/badge.svg)](https://github.com/EIDOSLAB/torchstain/actions)
+[![Full Tests](https://github.com/EIDOSLAB/torchstain/actions/workflows/tests_full.yml/badge.svg)](https://github.com/EIDOSLAB/torchstain/actions/workflows/tests_full.yml)
 [![Pip Downloads](https://img.shields.io/pypi/dm/torchstain?label=pip%20downloads&logo=python)](https://pypi.org/project/torchstain/)
 [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.7692014.svg)](https://doi.org/10.5281/zenodo.7692014)
 
@@ -79,7 +79,7 @@ Runtimes using the Macenko algorithm using different backends. Metrics were calc
 - [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.
-- [4] Ivanov, Desislav et al. "Multi-target stain normalization for histology slides". arXiv (preprint). 2024.
+- [4] Ivanov, Desislav et al. "Multi-target stain normalization for histology slides". 2nd International Workshop on Medical Optical Imaging and Virtual Microscopy Image Analysis (MOVI 2024), MICCAI. 2024.
 
 ## Citing
 

setup.py

@@ -6,7 +6,7 @@
 
 setup(
     name='torchstain',
-    version='1.3.0',
+    version='1.4.0',
     description='Stain normalization tools for histological analysis and computational pathology',
     long_description=README,
     long_description_content_type='text/markdown',

torchstain/base/normalizers/__init__.py

@@ -1,3 +1,4 @@
 from .he_normalizer import HENormalizer
 from .macenko import MacenkoNormalizer
+from .multitarget import MultiMacenkoNormalizer
 from .reinhard import ReinhardNormalizer

torchstain/base/normalizers/multitarget.py

@@ -1,6 +1,10 @@
-def MultiMacenkoNormalizer(backend='torch', **kwargs):
-    if backend == 'torch':
-        from torchstain.torch.normalizers.multitarget import MultiMacenkoNormalizer
-        return MultiMacenkoNormalizer(**kwargs)
+def MultiMacenkoNormalizer(backend="torch", **kwargs):
+    if backend == "numpy":
+        raise NotImplementedError("MultiMacenkoNormalizer is not implemented for NumPy backend")
+    elif backend == "torch":
+        from torchstain.torch.normalizers import TorchMultiMacenkoNormalizer
+        return TorchMultiMacenkoNormalizer(**kwargs)
+    elif backend == "tensorflow":
+        raise NotImplementedError("MultiMacenkoNormalizer is not implemented for TensorFlow backend")
     else:
-        raise Exception(f'Unsupported backend {backend}')
+        raise Exception(f"Unsupported backend {backend}")

torchstain/torch/normalizers/__init__.py

@@ -1,3 +1,3 @@
 from torchstain.torch.normalizers.macenko import TorchMacenkoNormalizer
-from torchstain.torch.normalizers.multitarget import MultiMacenkoNormalizer
+from torchstain.torch.normalizers.multitarget import TorchMultiMacenkoNormalizer
 from torchstain.torch.normalizers.reinhard import TorchReinhardNormalizer

torchstain/torch/normalizers/multitarget.py

@@ -1,16 +1,17 @@
 import torch
 from torchstain.torch.utils import cov, percentile
+
 """
 Implementation of the multi-target normalizer from the paper: https://arxiv.org/pdf/2406.02077
 """
-class MultiMacenkoNormalizer:
-    def __init__(self, norm_mode='avg-post'):
+class TorchMultiMacenkoNormalizer:
+    def __init__(self, norm_mode="avg-post"):
         self.norm_mode = norm_mode
         self.HERef = torch.tensor([[0.5626, 0.2159],
                                    [0.7201, 0.8012],
                                    [0.4062, 0.5581]])
         self.maxCRef = torch.tensor([1.9705, 1.0308])
-        self.updated_lstsq = hasattr(torch.linalg, 'lstsq')
+        self.updated_lstsq = hasattr(torch.linalg, "lstsq")
 
     def __convert_rgb2od(self, I, Io, beta):
         I = I.permute(1, 2, 0)
@@ -59,15 +60,15 @@ def __compute_matrices_single(self, I, Io, alpha, beta):
         return HE, C, maxC
 
     def fit(self, Is, Io=240, alpha=1, beta=0.15):
-        if self.norm_mode == 'avg-post':
+        if self.norm_mode == "avg-post":
             HEs, _, maxCs = zip(*(
                 self.__compute_matrices_single(I, Io, alpha, beta)
                 for I in Is
             ))
 
             self.HERef = torch.stack(HEs).mean(dim=0)
             self.maxCRef = torch.stack(maxCs).mean(dim=0)
-        elif self.norm_mode == 'concat':
+        elif self.norm_mode == "concat":
             ODs, ODhats = zip(*(
                 self.__convert_rgb2od(I, Io, beta)
                 for I in Is
@@ -83,7 +84,7 @@ def fit(self, Is, Io=240, alpha=1, beta=0.15):
             maxCs = torch.stack([percentile(C[0, :], 99), percentile(C[1, :], 99)])
             self.HERef = HE
             self.maxCRef = maxCs
-        elif self.norm_mode == 'avg-pre':
+        elif self.norm_mode == "avg-pre":
             ODs, ODhats = zip(*(
                 self.__convert_rgb2od(I, Io, beta)
                 for I in Is
@@ -100,7 +101,7 @@ def fit(self, Is, Io=240, alpha=1, beta=0.15):
             maxCs = torch.stack([percentile(C[0, :], 99), percentile(C[1, :], 99)])
             self.HERef = HE
             self.maxCRef = maxCs
-        elif self.norm_mode == 'fixed-single' or self.norm_mode == 'stochastic-single':
+        elif self.norm_mode == "fixed-single" or self.norm_mode == "stochastic-single":
             # single img
             self.HERef, _, self.maxCRef = self.__compute_matrices_single(Is[0], Io, alpha, beta)
         else:
@@ -127,4 +128,4 @@ def normalize(self, I, Io=240, alpha=1, beta=0.15, stains=True):
             E[E > 255] = 255
             E = E.T.reshape(h, w, c).int()
 
-        return Inorm, H, E
+        return Inorm, H, E