Pith. sign in

REVIEW 3 major objections 5 minor 31 references

Efficient Brain Imaging Analysis for Alzheimer's and Dementia Detection Using Convolution-Derivative Operations

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that Sobel gradient angle differences (SKAD) match Jacobian maps for dementia detection while cutting per-image computation by about 6.3x.

desk verdict SKAD is a cheap, plausible input feature for dementia MRI, but the paper overclaims that an intensity-gradient angle measures volume change; the empirical case is promising but needs statistical rigor and validation. read the letter →

arxiv 2411.13490 v2 pith:NIAE6GZZ submitted 2024-11-20 eess.IV cs.CVcs.NEcs.PF

Yasmine Mustafa , Mohamed Elmahallawy , Tie Luo This is my paper · ORCID
classification eess.IVcs.CVcs.NEcs.PF
keywords Alzheimer'sdiseasedementiaJacobianmapsSobelkernelvoxel-basedmorphometrystructuralMRIconvolutionatrophydetection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes Sobel Kernel Angle Difference (SKAD) as a cheap stand-in for Jacobian maps in voxel-based morphometry. SKAD computes the angle between 3D Sobel gradients of a registered patient MRI and the MNI template, and uses that angle difference as a signal of regional brain-volume change. The authors report that SKAD runs in 93 ms per image versus 584 ms for a Jacobian map (about 6.3x faster, and roughly 88x fewer floating-point operations), while a VGG-16 classifier trained on SKAD inputs reaches comparable accuracy to one trained on Jacobian maps: 91.06% on ADNI and 94.3% on OASIS-3. If it holds up, this gives neuroimaging pipelines a much cheaper preprocessing step for large-scale dementia screening.

What carries the argument

The central object is the Sobel kernel angle difference (SKAD), computed in Eq. 7 as $\left|\tan^{-1}\left(\frac{\sqrt{\nabla M_x^2 + \nabla M_y^2}}{\nabla M_z}\right) - \tan^{-1}\left(\frac{\sqrt{\nabla F_x^2 + \nabla F_y^2}}{\nabla F_z}\right)\right|$, where $\nabla M$ and $\nabla F$ are the 3D Sobel gradient vectors of the registered moving image and the fixed template. The 3x3x3 Sobel kernels replace explicit finite-difference derivatives with convolution, and the angle between gradient vectors is used as a scalar summary of local structural orientation disparity at each voxel. The machinery's job is to produce a single map per brain that highlights regions where the patient's image and the template disagree in gradient orientation, which the paper argues marks atrophy.

What would settle it

Compute both the SKAD map and the Jacobian determinant map from the same set of registered MRI scans and measure the voxelwise correlation between SKAD values and log-Jacobian determinants. If the correlation is near zero, or if SKAD's classification performance comes primarily from template registration artifacts rather than from true volume change, then the claim that SKAD captures regional volume variations is falsified.

Watch

Extended reading notes

Core claim

SKAD is defined by Eq. 7 as the absolute difference between the 3D gradient-direction angles of the warped patient image and the fixed MNI template, where gradients are obtained by convolving the image volume with 3x3x3 Sobel kernels in the x, y, and z directions. The paper's central claim is that this angle difference captures regional volume variations and signifies abnormalities or atrophy, making it a viable substitute for the Jacobian determinant map, which is the standard measure of local volume change during nonlinear registration. On the ADNI and OASIS-3 datasets, a VGG-16 model fed SKAD maps achieves 91.06% and 94.3% accuracy respectively, compared with 94.0% and 89.4% for Jacobian maps in the paper's own ablation, while SKAD costs 93 ms and 1.66 billion FLOPs versus 584 ms and 146 billion FLOPs for the Jacobian map. The authors conclude that SKAD is an efficient and competitive approach for dementia diagnosis.

Load-bearing premise

The load-bearing premise is that the difference in angle between the Sobel gradient orientations of a warped patient image and the fixed healthy template is a valid indicator of local brain-volume loss, even though the paper does not derive or empirically validate the link between this angle difference and the Jacobian determinant, which is the actual measure of volume change.

Editorial extensions

If this is right

  • A VGG-16 model trained on SKAD inputs reaches 91.06% accuracy on ADNI and 94.3% on OASIS-3, on par with Jacobian inputs in the paper's experiments.
  • Per-image preprocessing drops from 584 ms to 93 ms and from about 146 billion to 1.66 billion floating-point operations, making large-scale and repeated imaging analysis much cheaper.
  • Because SKAD has lower memory use (1199.7 MiB vs 1267.3 MiB for Jacobian maps), it can be applied to larger batches or on more modest hardware.
  • The method is not tied to Alzheimer's specifically and could extend to other neurodegenerative diseases where atrophy patterns matter, such as Parkinson's, if relevant imaging data are available.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct voxelwise comparison of SKAD against log-Jacobian determinants would clarify whether SKAD truly measures volume change or instead captures edge and boundary shifts; the paper does not report such a comparison.
  • Because SKAD only compares gradient orientations, it is insensitive to the magnitude of tissue loss; a region that shrinks while preserving its edge orientation could produce little or no SKAD signal, which would limit sensitivity in early atrophy.
  • SKAD's speed makes it a candidate preprocessing step for large self-supervised models that need many images; the paper notes this possibility but does not test it.
  • The angle difference could be computed earlier, directly from the deformation field, which might isolate the orientation-change component; that variant is not explored in the paper.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes Sobel Kernel Angle Difference (SKAD) as a computationally cheaper alternative to Jacobian maps for quantifying brain volume changes in dementia classification. SKAD is computed by convolving the registered patient MRI and the MNI template with 3D Sobel kernels, then taking the absolute difference of the polar angle formed by the three gradient components (Eq. 7). The authors evaluate SKAD versus registered images and Jacobian maps as inputs to a fine-tuned VGG-16 on ADNI and OASIS-3, report time and FLOP savings (93 ms vs 584 ms, 1.66B vs 146B FLOPs), and compare against prior Jacobian-based works.

Significance. If the central claim were established, the efficiency gain (84% time reduction, 99% FLOP reduction) would be practically useful for large-scale neuroimaging pipelines. The paper is credit-worthy in several respects: the Sobel kernels and angle formula are fixed a priori with no fitted constants, the within-paper ablation across registered, Jacobian, and SKAD inputs is honest, the subject-disjoint split is explicitly stated, and the profiling numbers are concrete and falsifiable. However, the conceptual foundation of the method — the claim that Eq. 7 captures regional volume variations — is not validated, and the empirical evaluation lacks statistical rigor. The significance of the work therefore depends on a load-bearing assumption that the manuscript does not yet support.

major comments (3)
  1. [Section III-C, Eq. (7)] The central claim that the angle difference between Sobel gradients of the warped image and the template signifies atrophy or volume change is unsupported. For an intensity-preserving warp, ∇(M∘φ) = Dφ^T (∇M)∘φ, so the angle depends on the full local deformation gradient and on the local intensity structure, not solely on the Jacobian determinant. In anatomically uniform regions the gradient magnitude can approach zero, making the ratio (√(∇Mx²+∇My²))/∇Mz unstable, and the absolute difference of arctangents does not handle the full 3D orientation (it reduces to the polar angle from the z-axis). No derivation, no voxel-wise comparison with det(J), and no empirical correlation is provided. Since the efficiency argument rests on SKAD being a substitute for Jacobian maps, please add a mathematical or empirical validation, for example a scatter plot or correlation coefficient between SKAD values and det(J) on a held-out set.
  2. [Section IV, Tables II and III] The empirical evaluation does not report error bars, confidence intervals, or statistical tests. Table II compares SKAD with prior works that use different models (CNN, RF+CNN), different numbers of input modalities (MRI only, MRI+CT, MRI+clinical data), and different datasets (ADNI vs OASIS-3), so the reported accuracies are not matched comparisons. The within-paper ablation in Table III is more informative, but without repeated runs or cross-validation folds it is impossible to know whether the 1–3 point gaps between Jacobian and SKAD are meaningful. Please report mean ± standard deviation over at least 5 runs or cross-validation folds, and restrict claims of competitiveness to the controlled within-paper comparison.
  3. [Section IV-A] The data split and augmentation protocol are incompletely specified. The rebalancing to (1233, 1233, 1233, 1233) for OASIS and (133, 133, 133, 133) for ADNI is described for the training set, but it is not stated whether the test set is augmented, whether the reported metrics are on the original test set, or how many unique subjects are in each split. Since augmentation can leak into test performance if applied after splitting, please state explicitly that augmentation is applied only to training images after the subject-disjoint split, and report the number of subjects (not scans) per class in train and test sets.
minor comments (5)
  1. [Abstract and Section III-C] The phrase "captures regional volume variations" is used before the connection to volume change is established; consider rewording to "gradient orientation disparity" pending the validation requested above.
  2. [Section IV-C, Table I] The 6.3x speedup compares SKAD (93 ms) with the Jacobian map (584 ms) in isolation, but the full pipeline still requires the 12,972 ms registration step. Please clarify in the text that the speedup applies only to the post-registration volumetric-change computation, not to the end-to-end pipeline.
  3. [Figure 6] The claim that SKAD "demonstrates the most stability" is based on visual inspection of training curves; reporting variance across runs would make this quantitative.
  4. [Throughout] Minor typographical issues include "ADIN" in the Figure 6 caption, "NVDIA" in Section IV-C, and a missing period between two sentences in the abstract.
  5. [References] Reference [6] appears as "V oxel-based morphometry" and should be corrected to "Voxel-based morphometry."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SKAD is defined by fixed Sobel convolutions and a fixed angle formula; the Jacobian-equivalence claim is an empirical assertion, not a construction.

full rationale

The paper's central chain is: register each subject to the MNI template with ANTs, define SKAD in Eq. 7 as the absolute difference of 3D gradient angles computed with fixed 3x3x3 Sobel kernels, feed the SKAD maps to a standard VGG-16 classifier, and compare accuracy and timing against Jacobian maps. No parameter in Eq. 7 is fitted to the classification outcome, and no term in SKAD is defined via the Jacobian determinant or via the disease labels. The assertion that the angle difference 'signifies abnormalities or atrophy' and the implicit claim that it proxies det(J) are not derived from first principles—the angle depends on the full local deformation gradient and on the image intensity structure, not on volume change alone, and it is undefined in flat regions. However, this is a correctness/validity gap rather than circularity: the paper does not define SKAD in terms of the Jacobian map, nor does it fit SKAD to reproduce Jacobian outputs. The efficiency comparison (93 ms vs 584 ms; 1.66 billion vs 146 billion FLOPs) is a direct profiling measurement from Table I and Fig. 5. Self-citations [22, 23] are prior works that used Jacobian maps and serve only as comparison baselines in Table II; they are not load-bearing for SKAD's core premise. The classification accuracy comparison is an independent empirical test, not a quantity forced by construction. Therefore the derivation chain is self-contained and no circularity is present.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

SKAD contains no fitted constants: the Sobel kernels and the angle formula are fixed. The paper relies on standard registration, standard derivative approximations, a clinical labeling scheme, and one unproven assumption: that gradient-angle differences represent volume changes. No new physical or computational entities are introduced.

assumptions (5)
  • domain assumption Nonlinear registration produces a deformation field that accurately maps each subject's anatomy to the MNI 152 template (Section III-A).
    Both Jacobian maps and SKAD are computed after ANTs registration; if registration is inaccurate, neither method reflects true anatomical differences.
  • ad hoc to paper Gradient orientation difference between the warped intensity image and the fixed template is a proxy for local volume change (Section III-C, Eq. 7).
    This premise is asserted without derivation or validation; it is the load-bearing step that makes SKAD a putative Jacobian alternative.
  • standard math 3D Sobel kernels approximate first-order spatial derivatives (Eqs. 4-6).
    Standard finite-difference derivative approximation from image processing, cited in [2, 11].
  • domain assumption VGG-16, pretrained on ImageNet, transfers to 3D MRI volumes (Section IV-B).
    The paper gives no detail on how the 2D architecture is adapted to 3D volumes, yet the model's accuracy is the empirical evidence.
  • domain assumption CDR-based labels and the four-class grouping (CN, MCI, MLD, SEV) are reliable ground truth (Section IV-A).
    All accuracy numbers depend on this clinical labeling.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Brain Imaging Analysis for Alzheimer's and Dementia Detection Using Convolution-Derivative Operations." pith.science (2026). https://pith.science/paper/NIAE6GZZ

@misc{pith2026241113490,
  author       = {Pith},
  title        = {Pith review of: Efficient Brain Imaging Analysis for Alzheimer's and Dementia Detection Using Convolution-Derivative Operations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NIAE6GZZ}},
  note         = {Machine review of arXiv:2411.13490}
}
read the original abstract

Alzheimer's disease (AD) is characterized by progressive neurodegeneration and results in detrimental structural changes in human brains. Detecting these changes is crucial for early diagnosis and timely intervention of disease progression. Jacobian maps, derived from spatial normalization in voxel-based morphometry (VBM), have been instrumental in interpreting volume alterations associated with AD. However, the computational cost of generating Jacobian maps limits its clinical adoption. In this study, we explore alternative methods and propose Sobel kernel angle difference (SKAD) as a computationally efficient alternative. SKAD is a derivative operation that offers an optimized approach to quantifying volumetric alterations through localized analysis of the gradients. By efficiently extracting gradient amplitude changes at critical spatial regions, this derivative operation captures regional volume variations Evaluation of SKAD over various medical datasets demonstrates that it is 6.3x faster than Jacobian maps while still maintaining comparable accuracy. This makes it an efficient and competitive approach in neuroimaging research and clinical practice.

Figures

Figures reproduced from arXiv: 2411.13490 by the authors.

Figure 1
Figure 1. Visualization of a deformation field from 6 view angles through deforming overlaid grids. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Schematic illustration of the 3D 3×3 Sobel kernels used in this work. These kernels are applied to both the moving and fixed images to perform convolutions in the x, y, and z directions. The resulting gradients are then used to compute the 3D angle difference. C. Sobel Kernel Angle Difference (SKAD) We propose Sobel Kernel Angle Difference (SKAD) as a novel method for characterizing brain volumetric changes. SKAD of… view at source ↗
Figure 3
Figure 3. Visualization of MRI-derived heat maps depicting the scans after registration, Jacobian Maps, and SKAD. Four cognitive stages are [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The MRI preprocessing pipeline depicted across three orthog [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Comparison of Floating-Point Operations (note the log scale). [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Multiple performance metrics for 3 methods on ADIN and OASIS datasets over 50 epochs. The text annotation describes SKAD. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 30 canonical work pages

  1. [1]

    Transformed domain convolutional neural network for alzheimer’s disease diagnosis using structural mri

    S Qasim Abbas, Lianhua Chi, and Yi-Ping Phoebe Chen. Transformed domain convolutional neural network for alzheimer’s disease diagnosis using structural mri. Pattern Recognition, 133:109031, 2023

    work page 2023

  2. [2]

    3d seismic edge detection using magic squares and cubes

    Peter Adetokunbo, Abdullatif A Al-Shuhail, and Saleh Al- Dossary. 3d seismic edge detection using magic squares and cubes. Interpretation, 4(3):T271–T280, 2016

    work page 2016

  3. [3]

    Preclinical stage alzheimer’s disease detec- tion using magnetic resonance image scans

    Fatih Altay et al. Preclinical stage alzheimer’s disease detec- tion using magnetic resonance image scans. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 15088–15097, 2021

    work page 2021

  4. [4]

    Memory_Profiler

    Altendorf et al. Memory_Profiler. https://github.com/pytho nprofilers/memory_profiler, 2024

    work page 2024

  5. [5]

    Detecting salt domes using a dip guided 3d sobel seismic attribute

    Ahmed Adnan Aqrawi, Trond Hellem Boe, and Sergio Bar- ros. Detecting salt domes using a dip guided 3d sobel seismic attribute. In SEG International Exposition and Annual Meet- ing, pages SEG–2011. SEG, 2011

    work page 2011

  6. [6]

    V oxel-based morphome- try—the methods

    John Ashburner and Karl J Friston. V oxel-based morphome- try—the methods. Neuroimage, 11(6):805–821, 2000

    work page 2000

  7. [7]

    Advanced normalization tools (ants)

    Brian B Avants et al. Advanced normalization tools (ants). Insight j, 2(365):1–35, 2009

    work page 2009

  8. [8]

    An explainable convolutional neural network for the early diagnosis of alzheimer’s disease from 18f-fdg pet

    Lisa Anita De Santi, Elena Pasini, Maria Filomena Santarelli, Dario Genovesi, and Vincenzo Positano. An explainable convolutional neural network for the early diagnosis of alzheimer’s disease from 18f-fdg pet. Journal of Digital Imaging, 36(1):189–203, 2023

    work page 2023

Show all 31 references
  1. [9]

    Brain templates and atlases

    Alan C Evans, Andrew L Janke, D Louis Collins, and Sylvain Baillet. Brain templates and atlases. Neuroimage, 62(2):911– 922, 2012

    2012

  2. [10]

    The clinical use of struc- tural mri in alzheimer disease

    Giovanni B Frisoni, Nick C Fox, Clifford R Jack Jr, Philip Scheltens, and Paul M Thompson. The clinical use of struc- tural mri in alzheimer disease. Nature Reviews Neurology, 6(2):67–77, 2010

    2010

  3. [11]

    Digital image processing: Pearson prentice hall

    Rafael C Gonzalez and Richard E Woods. Digital image processing: Pearson prentice hall. Upper Saddle River, NJ , 1(376-376):97, 2008

    2008

  4. [12]

    A voxel-based morphometric study of ageing in 465 normal adult human brains

    Catriona D Good, Ingrid S Johnsrude, John Ashburner, Richard NA Henson, Karl J Friston, and Richard SJ Frack- owiak. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage, 14(1):21–36, 2001

    2001

  5. [13]

    Un- derwater man-made object prediction using line detection technique

    Syed Safdar Hussain and Syed Sajjad Haider Zaidi. Un- derwater man-made object prediction using line detection technique. In Proceedings of the 2014 6th International Con- ference on Electronics, Computers and Artificial Intelligence (ECAI), pages 1–6. IEEE, 2014

    2014

  6. [14]

    Deep learning ap- proaches for covid-19 detection based on chest x-ray images

    Aras M Ismael and Abdulkadir ¸ Sengür. Deep learning ap- proaches for covid-19 detection based on chest x-ray images. Expert Systems with Applications, 164:114054, 2021

    2021

  7. [15]

    Deep learning based detection and analysis of covid- 19 on chest x-ray images

    Rachna Jain, Meenu Gupta, Soham Taneja, and D Jude He- manth. Deep learning based detection and analysis of covid- 19 on chest x-ray images. Applied Intelligence , 51:1690– 1700, 2021

    2021

  8. [16]

    Mark Jenkinson, Christian F Beckmann, Timothy EJ Behrens, Mark W Woolrich, and Stephen M Smith. Fsl. Neuroimage, 62(2):782–790, 2012

    2012

  9. [17]

    Design of an image edge detection filter using the so- bel operator

    Nick Kanopoulos, Nagesh Vasanthavada, and Robert L Baker. Design of an image edge detection filter using the so- bel operator. IEEE Journal of solid-state circuits, 23(2):358– 367, 1988

    1988

  10. [18]

    The Linux programming interface: a Linux and UNIX system programming handbook

    Michael Kerrisk. The Linux programming interface: a Linux and UNIX system programming handbook. No Starch Press, 2010

    2010

  11. [19]

    Structural consequences of diffuse traumatic brain injury: a large deformation tensor-based morphometry study

    Junghoon Kim, Brian Avants, Sunil Patel, John Whyte, Branch H Coslett, John Pluta, John A Detre, and James C Gee. Structural consequences of diffuse traumatic brain injury: a large deformation tensor-based morphometry study. Neuroimage, 39(3):1014–1026, 2008

    2008

  12. [20]

    Oasis-3: longitudinal neuroimaging, clin- ical, and cognitive dataset for normal aging and alzheimer disease

    Pamela J LaMontagne, Tammie LS Benzinger, John C Mor- ris, Sarah Keefe, Russ Hornbeck, Chengjie Xiong, Eliza- beth Grant, Jason Hassenstab, Krista Moulder, Andrei G Vlassenko, et al. Oasis-3: longitudinal neuroimaging, clin- ical, and cognitive dataset for normal aging and alz...

    2019

  13. [21]

    Correlation-aware coarse-to-fine mlps for deformable med- ical image registration

    Mingyuan Meng, Dagan Feng, Lei Bi, and Jinman Kim. Correlation-aware coarse-to-fine mlps for deformable med- ical image registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 9645–9654, 2024

    2024

  14. [22]

    Diagnosing Alzheimer’s disease using early-late multimodal data fusion with Jacobian maps

    Yasmine Mustafa and Tie Luo. Diagnosing Alzheimer’s disease using early-late multimodal data fusion with Jacobian maps. In IEEE International Conference on E-health Net- working, Application, and Services (Healthcom) , pages 49– 55, Dec. 2023

    2023

  15. [23]

    Unmasking dementia detec- tion by masking input gradients: A JSM approach to model interpretability and precision

    Yasmine Mustafa and Tie Luo. Unmasking dementia detec- tion by masking input gradients: A JSM approach to model interpretability and precision. In Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), pages 75– 90, 2024

    2024

  16. [24]

    Vit-ae++: improving vision transformer autoencoder for self-supervised medical image representations

    Chinmay Prabhakar, Hongwei Li, Jiancheng Yang, Suprosanna Shit, Benedikt Wiestler, and Bjoern Menze. Vit-ae++: improving vision transformer autoencoder for self-supervised medical image representations. In Medical Imaging with Deep Learning, pages 666–679. PMLR, 2024

    2024

  17. [25]

    The global prevalence of dementia: a systematic review and metaanal- ysis

    Martin Prince, Renata Bryce, Emiliano Albanese, Anders Wimo, Wagner Ribeiro, and Cleusa P Ferri. The global prevalence of dementia: a systematic review and metaanal- ysis. Alzheimer’s & dementia, 9(1):63–75, 2013

    2013

  18. [26]

    Fast robust automated brain extraction

    Stephen M Smith. Fast robust automated brain extraction. Human brain mapping, 17(3):143–155, 2002

    2002

  19. [27]

    A parameter-efficient deep learning approach to predict conversion from mild cognitive impairment to alzheimer’s disease

    Simeon Spasov, Luca Passamonti, Andrea Duggento, Pietro Lio, Nicola Toschi, Alzheimer’s Disease Neuroimaging Ini- tiative, et al. A parameter-efficient deep learning approach to predict conversion from mild cognitive impairment to alzheimer’s disease. Neuroimage, 189:276–287, 2019

    2019

  20. [28]

    The alzheimer’s disease neuroimaging initiative: progress report and future plans

    Michael W Weiner, Paul S Aisen, Clifford R Jack Jr, William J Jagust, John Q Trojanowski, Leslie Shaw, An- drew J Saykin, John C Morris, Nigel Cairns, Laurel A Beck- ett, et al. The alzheimer’s disease neuroimaging initiative: progress report and future plans. Alzheimer’s & De...

    2010

  21. [29]

    V oxel-based morphometry: an auto- mated technique for assessing structural changes in the brain

    Jennifer L Whitwell. V oxel-based morphometry: an auto- mated technique for assessing structural changes in the brain. Journal of Neuroscience, 29(31):9661–9664, 2009

    2009

  22. [30]

    A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia

    IC Wright, PK McGuire, J-B Poline, JM Travere, RM Murray, CD Frith, RSJ Frackowiak, and KJ Friston. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage, 2(4):244–252, 1995

    1995

  23. [31]

    Deeptag: An unsupervised deep learning method for motion tracking on cardiac tagging magnetic resonance images

    Meng Ye, Mikael Kanski, Dong Yang, Qi Chang, Zhennan Yan, Qiaoying Huang, Leon Axel, and Dimitris Metaxas. Deeptag: An unsupervised deep learning method for motion tracking on cardiac tagging magnetic resonance images. In Proceedings of the IEEE/CVF conference on computer visi...

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.