TABLE I.
Peer-reviewed journal articles depending on unreliable surrogates
| No. Publication | Summary of Problematic Methodology | |
|---|---|---|
| 1 | R. W. So, T. W. Tang, and A. C. Chung, “Non-rigid image registration of brain magnetic resonance images using graph-cuts,” Pattern Recognition, vol. 44, no. 10–11, pp. 2081–2092, 2011. | Claims “consistently higher registration accuracy” of proposed method based on tissue overlap scores, despite recovering known deformations. |
| 2 | S. Liao and A.C.S. Chung, “Feature based nonrigid brain MR image registration with symmetric alpha stable filters,” IEEE Transactions on Medical Imaging, vol. 29, no. 1, pp. 106–119, 2010. | Uses tissue overlap scores to compare new vs. existing registration algorithms. |
| 3 | A.M. Siddiqui, A. Masood, and M. Saleem, “A locally constrained radial basis function for registration and warping of images,” Pattern Recognition Letters, vol. 30, no. 4, pp. 377–390, 2009. | Compares transformation models using post-registration values of CC, MSD, and MI similarity measures. |
| 4 | H. Nam, R.A. Renaut, K. Chen, H. Guo, and G.E. Farin, “Improved inter-modality image registration using normalized mutual information with coarse-binned histograms,” Communications in Numerical Methods in Engineering, vol. 25, no. 6, pp. 583–595, 2009. | Applies known deformations, but then only uses intensity L2 error for comparison of registration results. |
| 5 | J. Larrey-Ruiz, R. Verdú-Monedero, and J. Morales-Sánchez, “A fourier domain framework for variational image registration,” Journal of Mathematical Imaging and Vision, vol. 32, no. 1, pp. 57–72, 2008. | Applies known deformations, but then only uses intensity PSNR, MI, and CR for comparison of registration results. |
| 6 | P. Zhilkin, M.E. Alexander, and J. Sun, “Nonlinear registration using variational principle for mutual information,” Pattern Recognition, vol. 41, no. 8, pp. 2493–2502, 2008. | Uses MSD and MI and measures of registration accuracy. |
| 7 | C. Frohn-Schauf, S. Henn, and K. Witsch, “Multigrid based total variation image registration,” Computing and Visualization in Science, vol. 11, no. 2, pp. 101–113, 2008. | Compares registration methods based on post-registration L2 image differences. |
| 8 | D.C. Paquin, D. Levy, and L. Xing, “Multiscale deformable registration of noisy medical images,” Mathematical Biosciences and Engineering, vol. 5, no. 1, pp. 125–144, 2008. | Uses post-registration correlation coefficient to “demonstrate the accuracy” of the proposed registration method. |
| 9 | D.C. Paquin, D. Levy, and L. Xing, “Hybrid landmark and multiscale deformable registration,” Mathematical Biosciences and Engineering, vol. 4, no. 4, pp. 711–737, 2007. | Uses difference images, CC and MSD to demonstrate registration “accuracy,” even though in one example 20 landmarks are used to drive the registration. |
| 10 | S. Tang and T. Jiang, “Nonrigid registration of medical image by linear singular blending techniques,” Pattern Recognition Letters, vol. 25, no. 4, pp. 399–405, 2004. | Uses post-registration SSD and MI to “evaluate […] accuracy” of proposed registration method against others. |
| 11 | B.C. Vemuri, J. Ye, Y. Chen, and C.M. Leonard, “Image registration via level-set motion: Applications to atlas-based segmentation,” Medical Image Analysis, vol. 7, no. 1, pp. 1–20, 2003. | Shows “difference image between evolved/transformed source image and the target image as a qualitative measure of the accuracy of the registration algorithm”; also uses CC to compare with other registration methods. |
PSNR = Peak Signal to Noise Ratio; CR = Correlation Ratio; CC = Correlation Coefficient; SSD = Sum of Squared Differences; MI = Mutual Information