Similarity and Quality Metrics for MR Images

This repository provides the metrics and distortions, that were presented in the paper:

M. Dohmen, M. Klemens, T. Truong, I. Baltruschat, M. Lenga: "Similarity and Quality Metrics for MR Image-to-Image Translation" submitted to Nature Scientific Reports (see preprint)

Installation

Clone repository

cd into iml-medical-image-metrics
pip install medimetrics

Example Usage

import numpy as np

from medimetrics.metrics import MSE, BlurEffect
from medimetrics.distortions import GaussianBlur, GammaHigh


blur = GaussianBlur(max_strength=5)
darker = GammaHigh(max_strength=5)

image_true = np.linspace(0.0, 1.0, 128*128).reshape(128, 128)
image_test_blurred = blur(image_true, strength=5)
image_test_darker = darker(image_true, strength=5)

mse = MSE()
print(mse.compute(image_true, image_test_darker))
> 0.0555207631635955

blur_effect = BlurEffect()
print(blur_effect.compute(image_test_blurred))
> 0.9772262051422484

Metrics

Reference Metrics

• MSE: Mean Squared Error
• SSIM: Structural Similarity Index Measure
  • Wang, Z., Bovik, A. C., Sheikh, H. R. & Simoncelli, E. P. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Process. 13, 600–12 (2004).
  • Implementation adapted from torchmetrics
• MAE: Mean Absolute Error
• MS-SSIM: Multi-Scale SSIM
  • Wang, Z., Simoncelli, E. & Bovik, A. Multiscale structural similarity for image quality assessment. In The Thirty-Seventh Asilomar Conference on Signals, Systems and Computers, 2003, vol. 2, 1398–1402 Vol.2, DOI: 10.1109/ACSSC.2003.1292216 (2003).
  • Implementation adapted from torchmetrics
• CW-SSIM: Complex-Wavelet SSIM
  • Sampat, M. P., Wang, Z., Gupta, S., Bovik, A. C. & Markey, M. K. Complex wavelet structural similarity: A new image similarity index. IEEE Transactions on Image Process. 18, 2385–2401, DOI: 10.1109/TIP.2009.2025923 (2009).
  • Implementation adapted from github/dingkeyan93
• PSNR : Peak-Signal-to-Noise-Ratio
  • Korhonen, J. & You, J. Peak signal-to-noise ratio revisited: Is simple beautiful? In 2012 Fourth International Workshop on Quality of Multimedia Experience, 37–38, DOI: 10.1109/QoMEX.2012.6263880 (2012)
• LPIPS : Learned Perceptual Image Patch Similarity
  • Zhang, R., Isola, P., Efros, A. A., Shechtman, E. & Wang, O. The unreasonable effectiveness of deep features as a perceptual metric. CoRR ayn/1801.03924 (2018). 1801.03924.
  • Implementation integrated from pypi
• DISTS : Deep Image Stucture and Texture Similarity
  • Ding, K., Ma, K., Wang, S. & Simoncelli, E. P. Image quality assessment: Unifying structure and texture similarity. IEEE Transactions on Pattern Analysis Mach. Intell. 44, 2567–2581, DOI: 10.1109/TPAMI.2020.3045810 (2022).
  • Implementation adapted from github/dingkeyan93
• PCC : Pearson Correlation Coefficient
  • Implementation integrated from numpy
• NMI : Normalized Mutual Information
  • Maes, F., Collignon, A., Vandermeulen, D., Marchal, G. & Suetens, P. Multimodality image registration by maximization of mutual information. IEEE Transactions on Med. Imaging 16, 187–198, DOI: 10.1109/42.563664 (1997).
  • Implementation adapted from scikit-image

Non-Reference Metrics

• BlurWidths:
  • Marziliano, P., Dufaux, F., Winkler, S. & Ebrahimi, T. A no-reference perceptual blur metric. In Proceedings. International Conference on Image Processing, vol. 3, III–III, DOI: 10.1109/ICIP.2002.1038902 (2002).
  • Implementation inspired by: github/affaalfiandy
• BlurJNB:
  • Ferzli, R. & Karam, L. J. A no-reference objective image sharpness metric based on the notion of just noticeable blur (jnb). IEEE Transactions on Image Process. 18, 717–728, DOI: 10.1109/TIP.2008.2011760 (2009).
  • Implementation adapted from: github/davidatroberts
• BlurCPBD:
  • Narvekar, N. D. & Karam, L. J. A no-reference image blur metric based on the cumulative probability of blur detection (cpbd). IEEE Transactions on Image Process. 20, 2678–2683, DOI: 10.1109/TIP.2011.2131660 (2011).
  • Implementation adapted from: github/x64746b
• BlurEffect:
  • Crété-Roffet, F., Dolmiere, T., Ladret, P. & Nicolas, M. The Blur Effect: Perception and Estimation with a New No- Reference Perceptual Blur Metric. In SPIE Electronic Imaging Symposium Conf Human Vision and Electronic Imaging, vol. XII, EI 6492–16 (San Jose, United States, 2007).
  • Implementation adapted from scikit-image
• BlurRatio / MeanBlur:
  • Choi, M. G., Jung, J. H. & Jeon, J. W. No-reference image quality assessment using blur and noise. Int. J. Electr. Comput. Eng. 3, 184–188 (2009).
  • Own implementation
• BRISQUE:
  • Mittal, A., Moorthy, A. K. & Bovik, A. C. No-reference image quality assessment in the spatial domain. IEEE Transactions on Image Process. 21, 4695–4708, DOI: 10.1109/TIP.2012.2214050 (2012).
  • Implementation integrated from pypi
• NIQE: Natural Image Quality Estimator
  • Mittal, A., Soundararajan, R. & Bovik, A. C. Making a “completely blind” image quality analyzer. IEEE Signal Process. Lett. 20, 209–212, DOI: 10.1109/LSP.2012.2227726 (2013).
  • Implementation adapted from github/guptapraful
• VarLaplace: Variance of Laplacian Pech-Pacheco, J., Cristobal, G., Chamorro-Martinez, J. & Fernandez-Valdivia, J. Diatom autofocusing in brightfield microscopy: a comparative study. In Proceedings 15th International Conference on Pattern Recognition. ICPR-2000, vol. 3, 314–317 vol.3, DOI: 10.1109/ICPR.2000.903548 (2000).
• MeanTotalVar: Mean Total Variation
  • Rudin, L. I., Osher, S. & Fatemi, E. Nonlinear total variation based noise removal algorithms. Phys. D: Nonlinear Phenom. 60, 259–268, DOI: https://doi.org/10.1016/0167-2789(92)90242-F (1992)
  • Own implementation
• MLC / MSLC: Mean Line Correlation / Mean Shifted Line Correlation
  • Schuppert et al., C. Whole-body magnetic resonance imaging in the large population-based german national cohort study: Predictive capability of automated image quality assessment for protocol repetitions. Investig. Radiol. 57 (2022).
  • Own implementation

Segmentation

• DICE : DICE
  • Dice, L. R. Measures of the amount of ecologic association between species. Ecology 26, 297–302, DOI: https://doi.org/10.2307/1932409 (1945).
  • Own implementation

Distortions

• Stripes Scales up the intensity of a single point in k-space, which leads to stripe artifacts
• Bias Field Applies a synthetic bias field to the image
• Ghosting Applies 2 ghosts in y-direction to the image
• Gaussian Blur Applies a Gaussian filter to the image, which leads to blurring
• Gaussian Noise Adds Gaussian noise to the image
• Replace Artifact Mirrors a fraction of the upper half of the image to the lower half and replaces this part.
• Shift Intensity Adds a fraction of the data range to all pixels
• Gamma High Normalizes the image to range (0, 1). Takes all pixels to the power of gamma, with gamma > 1, and then scales back to the original data range. All image intensities become lower, the center of the histogram is effected the most.
• Gamma Low Normalizes the image to range (0, 1). Takes all pixels to the power of gamma, with gamma < 1, and then scales back to the original data range. All image intensities become higher, the center of the histogram is effected the most.
• Translation Shifts the image by a fraction of the image shape in x- and y-direction
• Elastic Deform Randomly displaces nodes of a grid on the image and linearly interpolates the image between the nodes, which creates an elastic deformation