Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

SPM 25: open source neuroimaging analysis software

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read The paper reports the release of SPM 25.01, the first openly developed version of the Statistical Parametric Mapping software in a decade, with new Bayesian and OPM-MEG methods and MATLAB-free access.

desk verdict SPM 25.01 is a real, verifiable release note that improves the field's standard neuroimaging software workflow, though the paper's validation evidence for newly added methods is thin. read the letter →

arxiv 2501.12081 v1 pith:XOUO4ELF submitted 2025-01-21 q-bio.QM q-bio.NC

classification q-bio.QMq-bio.NC
keywords statisticalparametricmappingneuroimagingsoftwareSPM25.01opensourceMEG/EEGanalysisOPM-MEGBayesianspectraldecompositionPythonwrapper
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

This paper announces SPM 25.01, the first public release of the Statistical Parametric Mapping software in a decade and the first developed openly on GitHub. It claims the release folds ten years of new methods into the long-tested MATLAB/C core: Bayesian Spectral Decomposition for M/EEG spectra, interference cancellation for OPM-MEG, SCOPE distortion correction, multi-brain normalisation, and updated Active Inference behavioural modelling. It also claims new open-science infrastructure: automated tests, automatically built standalone and container versions, a documentation website, and spm-python, a wrapper that will let users run SPM from Python without MATLAB. The significance is practical: SPM is one of the most widely used toolboxes for testing hypotheses about brain structure and function, so this release changes how its users can run analyses and what methods are available.

What carries the argument

The central object is SPM itself: an integrated MATLAB/C codebase that applies voxel-wise General Linear Models to neuroimaging data, haemodynamic response convolution models for fMRI, random field theory corrections, voxel-based morphometry, dynamic causal modelling, and M/EEG source localisation. The release wraps this long-tested core in a new open development and build pipeline with automated tests, containers, and calendar versioning, and extends it with named new tools such as Bayesian Spectral Decomposition (BSD), which treats the periodic and aperiodic components of neural spectra as parameters of a generative model with posterior estimates, and the OPM interference-cancellation algorithms that model environmental fields as homogeneous fields or adaptive multipole expansions.

What would settle it

Run SPM 25.01's Bayesian Spectral Decomposition on the spectra used in the BSD paper and its OPM interference-cancellation routines on the datasets used in the adaptive multipole paper, then compare the outputs with the reference implementations; a mismatch would falsify the claim that the new components are correctly implemented. Also run a standard fMRI GLM analysis on a public dataset and check that SPM 25.01 reproduces SPM 12 results to numerical precision.

Watch

Extended reading notes

Core claim

The central claim is that SPM 25.01 is a stable, major new version of an established open-source neuroimaging analysis package that incorporates ten years of methodological developments. The paper describes the release as integrating novel analysis methods, such as Bayesian Spectral Decomposition (BSD) for parameterising neural power spectra with formal statistics, OPM interference cancellation via homogeneous field correction and adaptive multipole models, SCOPE for susceptibility-distortion correction, and multi-brain spatial normalisation, into the existing SPM framework of generative models and parametric statistics. It further claims that the software is now developed openly on a public repository platform, with automated regression tests, a new documentation website, automatically built standalone and container versions, and an in-development spm-python wrapper that will make SPM accessible from Python without a MATLAB license.

Load-bearing premise

The release's promise depends on the newly added routines, such as BSD, OPM interference cancellation, SCOPE, multi-brain normalisation, and the spm-python wrapper, being correctly implemented, because the paper gives no test results, benchmarks, or example outputs for them.

Editorial extensions

If this is right

  • A researcher without a MATLAB licence can run SPM 25.01 analyses from the command line via the automatically built standalone, or inside Docker or Singularity containers.
  • When spm-python ships in the first quarter of 2025, Python code will be able to call SPM's MATLAB/C routines directly, without a MATLAB installation.
  • OPM-MEG users gain multi-manufacturer file IO, array simulation, and two interference-cancellation methods, supporting studies that use wearable, moving-head MEG.
  • Users of behavioural modelling gain the latest Active Inference tooling, including POMDP inversion schemes, continuous active filtering, and hierarchical compositions.
  • MRI pipelines gain population-shaped normalisation through the Multi-Brain Toolbox and blip-up/blip-down distortion correction through SCOPE.

Reading between the lines

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

  • Beyond the paper, the move to a public development repository invites a direct test of community engagement: whether outside contributors submit improvements and whether the new regression tests catch platform-specific bugs that the old private development workflow missed.
  • Because the paper reports no test outputs or benchmarks for the new components, any independent replication of the cited BSD and OPM results using the released code would strengthen confidence beyond the paper's own assurances.
  • If spm-python delivers on its promise, SPM's MATLAB/C core could outlive the commercial viability of MATLAB itself, but the wrapper's success will depend on whether the compiled standalone runs reliably across diverse high-performance computing and cloud environments.
  • The BSD implementation may shift how spectral parameterisation is done in M/EEG studies, but that depends on whether researchers adopt the Bayesian formalism over the existing frequentist FOOOF approach; a head-to-head comparison on openly shared spectra would clarify the choice.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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. This paper announces the release of SPM 25.01, a major new version of the Statistical Parametric Mapping neuroimaging software package. It describes the move to a public GitHub repository, the introduction of automated tests and builds, a new documentation website, and lists several new analysis features in MRI, M/EEG, optically pumped magnetometer (OPM) processing, Bayesian statistics, and behavioural modelling. The paper also outlines a strategy for making SPM accessible without MATLAB, including a planned Python wrapper (spm-python) and the existing standalone and container distribution routes.

Significance. SPM is one of the most widely used software packages in neuroimaging, so a new release after more than a decade is intrinsically significant. If the claims are accurate, SPM 25.01 provides a stable, publicly developed release with new methods (Bayesian Spectral Decomposition, OPM interference correction, FOOOF/spectral parameterisation), automated build and container support, and a new documentation hub. The release itself is verifiable via the public GitHub repository, which is a concrete strength and aligns with open-science principles. However, the paper provides no release-specific evidence that the newly advertised methods are correctly implemented in SPM 25.01; the validation burden rests on external papers and on the long history of the MATLAB/C core, which does not cover these recent additions. The overall significance therefore depends on information that is not present in the manuscript.

major comments (3)
  1. [Sections 2.3.2 and 2.3.3] The paper lists Bayesian Spectral Decomposition (BSD), FOOOF (Specparam), and the OPM interference cancellation algorithms (Homogeneous Field Correction and Adaptive Multipole Models) as major new features of SPM 25.01, but provides no test results, example outputs, benchmarks, or pointers to reproducible runs for the SPM implementations. The blanket statement in Section 2.3.6 that the code has been 'highly optimised and thoroughly tested over 30 years of development' applies to the long-standing MATLAB/C core and cannot validate these relatively new additions. Because the central claim is that SPM 25.01 'incorporates novel analysis methods', this validation gap is load-bearing. Please add a section with results from the automated tests or Continuous Integration (CI), or provide a documented example dataset and expected outputs for each new feature, so that the implementations can be checked against the cited methods.
  2. [Section 2.3.6] The paper describes 'SPM without MATLAB' as part of the release strategy, but the Python wrapper spm-python is explicitly stated to be 'in the final stages of development and will be released in the first quarter of 2025', which is after SPM 25.01. The abstract and summary should make clear that Python support is a planned near-term addition and not a feature of the currently announced release. As written, readers could reasonably attribute Python access to SPM 25.01, which would be misleading and overstates the release's capabilities.
  3. [Section 2.1] The paper states that the move to GitHub has introduced 'automated unit and regression tests across platforms', but it does not provide any evidence that these tests exist, pass, or cover the new features. For a software release note, this claim should be substantiated with a link to the CI workflow (e.g., GitHub Actions status badge), a list of tested platforms, and a summary of test coverage or at least a list of the test suites. Without this, the 'improved practices for open science and software development' part of the central claim is not verifiable from the manuscript.
minor comments (5)
  1. [Section 2.3.3] The reference for Adaptive Multipole Models (Tierney et al., 2024) is given only as 'Wiley Online Library'; please provide full bibliographic details, including volume, article number, pages, or DOI.
  2. [Section 3] In addition to the GitHub releases page, please provide a persistent identifier (e.g., a Zenodo DOI) for the exact release version described, so that the software can be cited unambiguously in future studies.
  3. [Section 2.3.6] The statement that 'approximately 90% of the SPM 25.01 source code is written in MATLAB and the remainder is written in C' would benefit from a brief description of how this ratio was measured (e.g., line counts or file sizes) or a pointer to the relevant repository statistics.
  4. [Section 2.1] Please specify which platforms the automated tests are run on (e.g., Linux, macOS, Windows) and where the Continuous Integration status is publicly visible, so that users can independently confirm the claim of cross-platform testing.
  5. [Throughout] The manuscript alternates between 'SPM 25' and 'SPM 25.01' when referring to the release; please standardise the terminology to avoid confusion between the version series and the specific release.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: SPM 25.01 is a verifiable software release; citations to prior method papers, including same-group work, are not load-bearing reductions.

full rationale

SPM 25 is a software release note rather than a derivation or modeling paper: it contains no equations, no fitted parameters, and no quantities predicted from inputs. The central claim, that SPM 25.01 is released with the listed features, is externally checkable through the public GitHub repository, release artifacts, containers, and automated builds. The new capabilities (BSD, FOOOF integration, OPM interference methods) are delegated to cited method papers, including some by authors of this paper, but this is standard practice for a software announcement and not a circular reduction: the cited papers contain their own derivations and evaluations, and the implementations here are open source and independently runnable. The only forward-looking item (spm-python release in Q1 2025) is explicitly announced as not part of SPM 25.01, so it does not function as a claimed result dressed as a prediction. No step of the announcement reduces, by definition or by construction, to its own inputs. Score 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

No free parameters are fitted in this software release note. No new theoretical entities are proposed. The analysis rests on the body of cited neuroimaging methods and on the assumed reliability of the codebase.

assumptions (2)
  • domain assumption The cited source publications describe correct and valid methods.
    Section 2.3 lists features as implementations of (or inspired by) published works, e.g., FOOOF from Donoghue et al. (2020), BSD from Medrano et al. (2024), and OPM methods from Tierney et al. (2021, 2024); the paper does not re-validate them.
  • domain assumption The MATLAB/C codebase is reliable because of its long development history.
    Section 2.3.6 states the code 'has been highly optimised and thoroughly tested over 30 years of development' but provides no test reports or benchmarks to support this.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SPM 25: open source neuroimaging analysis software." pith.science (2026). https://pith.science/paper/XOUO4ELF

@misc{pith2026250112081,
  author       = {Pith},
  title        = {Pith review of: SPM 25: open source neuroimaging analysis software},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XOUO4ELF}},
  note         = {Machine review of arXiv:2501.12081}
}
read the original abstract

Statistical Parametric Mapping (SPM) is an integrated set of methods for testing hypotheses about the brain's structure and function, using data from imaging devices. These methods are implemented in an open source software package, SPM, which has been in continuous development for more than 30 years by an international community of developers. This paper reports the release of SPM 25.01, a major new version of the software that incorporates novel analysis methods, optimisations of existing methods, as well as improved practices for open science and software development.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. petBrain: A New Pipeline for Amyloid, Tau Tangles and Neurodegeneration Quantification Using PET and MRI

    eess.IV 2025-06 conditional novelty 5.0 of 10

    petBrain is a fully automated, web-based pipeline that computes Centiloid, CenTauRz, and HAVAs scores for A/T2/N staging, with validation against SPM, B-PIP, fluid biomarkers, and cognition on ADNI.

Reference graph

Works this paper leans on

27 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [1]

    L., Skare, S., & Ashburner, J

    Andersson, J. L., Skare, S., & Ashburner, J. (2003). How to corr ect susceptibility distortions in spin-echo echo-planar images: Application to diffusion tensor imaging . Neuroimage, 20 (2), 870–888

  2. [2]

    Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometr y—the methods. Neuroimage, 11 (6), 805–821

  3. [3]

    S ., Mu˜ noz, L

    Boto, E., Holmes, N., Leggett, J., Roberts, G., Shah, V., Meyer, S. S ., Mu˜ noz, L. D., Mullinger, K. J., Tierney, T. M., Bestmann, S., & others. (2018). Moving magneto encephalography towards real-world applications with a wearable system. Nature, 555 (7698), 657–661

  4. [4]

    Brudfors, M., Balbastre, Y., Flandin, G., Nachev, P., & Ashburner, J . (2020). Flexible bayesian modelling for nonlinear image registration. Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference , Lima, Peru, October 4–8, 2020,

  5. [5]

    J., Varma, P., Sebastian, P., G ao, R., Noto, T., Lara, A

    Donoghue, T., Haller, M., Peterson, E. J., Varma, P., Sebastian, P., G ao, R., Noto, T., Lara, A. H., Wallis, J. D., Knight, R. T., & others. (2020). Parameterizing neur al power spectra into periodic and aperiodic components. Nature Neuroscience, 23 (12), 1655–1665

  6. [6]

    J., Harrison, L., & Penny, W

    Friston, K. J., Harrison, L., & Penny, W. (2003). Dynamic causal mo delling. Neuroimage, 19 (4), 1273–1302

  7. [7]

    J., Holmes, A

    Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J.-P., Frith, C. D., & Frackowiak, R. S. (1994). Statistical parametric maps in functional imaging: A gener al linear approach. Human Brain Mapping, 2 (4), 189–210

  8. [8]

    J., Jezzard, P., & Turner, R

    Friston, K. J., Jezzard, P., & Turner, R. (1994). Analysis of funct ional MRI time-series. Human Brain Mapping, 1 (2), 153–171

Show all 27 references
  1. [9]

    Zeidman, P. (2016). Bayesian model reduction and empirical bayes for group (DCM) studies. Neuroimage, 128, 413–431

  2. [10]

    J., Parr, T., & Vries, B

    Friston, K. J., Parr, T., & Vries, B. de. (2017). The graphical brain : Belief propagation and active inference. Network Neuroscience, 1 (4), 381–414

  3. [11]

    J., Preller, K

    Friston, K. J., Preller, K. H., Mathys, C., Cagnan, H., Heinzle, J., Razi, A., & Zeidman, P. (2019). Dynamic causal modelling revisited. Neuroimage, 199, 730–744

  4. [12]

    J., Sajid, N., Quiroga-Martinez, D

    Friston, K. J., Sajid, N., Quiroga-Martinez, D. R., Parr, T., Price, C. J., & Holmes, E. (2021). Active listening. Hearing Research, 399, 107998

  5. [13]

    Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., & Pezzulo, G . (2017). Active inference: A process theory. Neural Computation , 29 (1), 1–49

  6. [14]

    Friston, K., & Frith, C. (2015). A duet for one. Consciousness and Cognition , 36, 390–405

  7. [15]

    Friston, K., Parr, T., & Zeidman, P. (2018). Bayesian model reduct ion. arXiv Preprint arXiv:1805.07092

  8. [16]

    J., & Zeidman, P

    Jafarian, A., Litvak, V., Cagnan, H., Friston, K. J., & Zeidman, P. (20 19). Neurovascular cou- pling: Insights from multi-modal dynamic causal modelling of fMRI and MEG. arXiv Preprint arXiv:1903.07478

  9. [17]

    Josephs, O., Turner, R., & Friston, K. (1997). Event-related f MR I. Human Brain Mapping , 5 (4), 243–248

  10. [18]

    A., Seymour, R

    Medrano, J., Alexander, N. A., Seymour, R. A., & Zeidman, P. (2024) . BSD: A bayesian framework for parametric models of neural spectra. arXiv Preprint arXiv:2410.20896

  11. [19]

    C., Barnes, G

    Walker, M. C., Barnes, G. R., & Vivekananda, U. (2024). Detection o f mesial temporal lobe epilepsy with OP-MEG. medRxiv, 2024–2010

  12. [20]

    M., Seymour, R

    Mellor, S., Tierney, T. M., Seymour, R. A., Timms, R. C., O’Neill, G. C., Alexa nder, N., Spedden, M. E., Payne, H., & Barnes, G. R. (2023). Real-time, model-based ma gnetic field correction for moving, wearable MEG. NeuroImage, 278, 120252

  13. [21]

    Parr, T., Pezzulo, G., & Friston, K. J. (2022). Active inference: The free energy principle in mind, brain, and behavior . MIT Press

  14. [22]

    Schwartenbeck, P., & Friston, K. (2016). Computational phenot yping in psychiatry: A worked example. Eneuro, 3 (4)

  15. [23]

    E., O’Neill, G

    Spedden, M. E., O’Neill, G. C., Tierney, T. M., West, T. O., Schmidt, M., Me llor, S., Farmer, S. F., Bestmann, S., & Barnes, G. R. (2024). Towards non-invasive ima ging through spinal-cord 5 generated magnetic fields. Frontiers in Medical Technology, 6, 1470970

  16. [24]

    M., Alexander, N., Mellor, S., Holmes, N., Seymour, R., O’Neill, G

    Tierney, T. M., Alexander, N., Mellor, S., Holmes, N., Seymour, R., O’Neill, G. C., Maguire, E. A., & Barnes, G. R. (2021). Modelling optically pumped magnetometer interference in MEG as a spatially homogeneous magnetic field. NeuroImage, 244, 118484

  17. [25]

    U., Dalal, S

    Westner, B. U., Dalal, S. S., Gramfort, A., Litvak, V., Mosher, J. C., O ostenveld, R., & Schoffelen, J.-M. (2022). A unified view on beamformers for m/EEG source reco nstruction. NeuroImage, 246, 118789

  18. [26]

    J., Marrett, S., Neelin, P., Vandal, A

    Worsley, K. J., Marrett, S., Neelin, P., Vandal, A. C., Friston, K. J., & E vans, A. C. (1996). A unified statistical approach for determining significant signals in imag es of cerebral activation. Human Brain Mapping , 4 (1), 58–73

  19. [27]

    Yan, Y., Balbastre, Y., Brudfors, M., & Ashburner, J. (2022). Fac torisation-based image labelling. Frontiers in Neuroscience, 15, 818604. 6

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.