{"id":"14d8ae15-c97a-466a-9482-f837cce104fb","arxiv_id":"2501.12081","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"SPM 25.01, the first GitHub-based release of the popular neuroimaging analysis package, adds new MRI, M/EEG, OPM, and Bayesian modeling tools and is now accessible from Python.","lead":"SPM 25.01 is a major update to a free, long-used toolbox for analyzing brain imaging data, adding new tools for MRI, M/EEG, and wearable magnetometers. The paper is a release announcement; its value is in making the updated open-source software and its new documentation available to the neuroimaging community.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Release itself is verifiable, but the paper asserts new BSD/OPM methods are incorporated without release-specific validation; Python access is explicitly forward-looking, not part of SPM 25.01.","rationale":"The reader's conditional verdict is appropriate and my read does not move it. The paper's main assertion—that SPM 25.01 exists as an open release—is easily checked and is backed by a public GitHub repository, automated build, containers, and standalone packaging; these are concrete artifacts that should be credited. The weakest point is genuinely the unshown correctness of newly listed methods, since 'thoroughly tested over 30 years' cannot cover BSD, OPM multipole correction, or the M/EEG beamformer changes. This is a correctness risk, not a proof gap, and it is mitigable by code inspection and running the public test suite. The Python wrapper is explicitly announced as a future release, so it should not be read as a delivered feature of SPM 25.01. No basis appears for REJECT or for stronger language; the open availability of the code is strong independent evidence that the central release claim is not empty. I therefore leave the verdict unchanged and agree with the reader's identification of the soft spot.","tokens_in":5251,"tokens_out":4234,"duration_ms":43598,"concrete_test":"Check out the git tag for the SPM 25.01 release, run the repository's automated unit and regression tests, and execute one BSD example in MEEGtools on a synthetic spectrum with known periodic and aperiodic parameters; compare the recovered parameters with the reference implementation in Medrano et al. (2024) and with FOOOF output. If the BSD posterior deviates beyond the reference benchmark error or the test suite fails in a clean MATLAB/Octave environment, the assertion that the new methods are correctly incorporated is weakened. Separately, confirm whether spm-python is present in the release; if it is absent, the Python-accessible claim should be labelled as planned, not current.","verdict_should_be":"UNCHANGED","load_bearing_attack":"To support the central claim that SPM 25.01 'incorporates novel analysis methods', the new components must be present and behave as described. Section 2.3.2 and 2.3.3 list Bayesian Spectral Decomposition, FOOOF, and OPM interference correction (HFC and Adaptive Multipole Models), each delegated to external citations (Medrano et al. 2024; Donoghue et al. 2020; Tierney et al. 2021, 2024), with no test results, example outputs, or comparisons shown for the SPM 25.01 implementations. The blanket statement in Section 2.3.6 that 'this code has been highly optimised and thoroughly tested over 30 years' applies to the MATLAB/C core, not to these relatively new additions, so it cannot carry the validation burden for them. In addition, the headline Python accessibility promise in Section 2.3.6 is explicitly not delivered in SPM 25.01 ('will be released in the first quarter of 2025'), making that part of the announcement forward-looking. This is a software-verification gap, not an internal inconsistency; because the repository is public and runnable, the gap can be closed by direct testing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","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.","tokens_in":5441,"tokens_out":5374,"duration_ms":54641,"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":[{"comment":"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.","section":"Sections 2.3.2 and 2.3.3"},{"comment":"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.","section":"Section 2.3.6"},{"comment":"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.","section":"Section 2.1"}],"minor_comments":[{"comment":"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.","section":"Section 2.3.3"},{"comment":"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.","section":"Section 3"},{"comment":"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.","section":"Section 2.3.6"},{"comment":"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.","section":"Section 2.1"},{"comment":"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.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"This is a software release note rather than a methods or validation paper. The core claim (that SPM 25.01 exists and is publicly available) is easily verifiable, and the move to GitHub is a positive step. However, the paper advertises several new methods without any release-specific evidence of correctness. The authors are strongly encouraged to link to their CI test results and to provide a minimal reproducible example for each new feature. The citation pattern leans heavily on the authors' own prior work, which is common in release notes but makes independent validation particularly important. If the authors can supply the requested validation evidence, the paper would be well within the scope of a software-focused neuroimaging journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"SPM 25.01 is a software release note, and the main thing to know is that the central claim is verifiable and true: the code is on GitHub, the release exists, and the packaging improvements (automated builds, containers, documentation site) are real workflow changes. The paper is not a methods paper; the novel analysis methods (BSD, FOOOF integration, OPM interference cancellation) are cited to prior papers, mostly from the same group. That is fine for a release note, but it does mean the release-specific validation burden is on the software itself, and the paper does not carry that burden. Section 2.3.6's \"highly optimised and thoroughly tested over 30 years\" applies to the MATLAB/C core, not to the new additions, and the paper shows no test outputs or comparisons for the new components. The stress-test note is right: that is a software-verification gap, not an internal contradiction. The Python wrapper is explicitly not in 25.01; the section is honest about that. For a field standard, a release note that moves development to public GitHub, adds automated tests and containers, and bundles previously validated methods is a legitimate and useful artifact. It deserves a serious referee: someone should check that the release actually exists, that the bundled toolboxes run, and that the citations match the implementations. The paper could be improved by adding a table of CI test results, a minimal example output per new feature, and a clear statement that Python is a future release. The citation pattern is heavy on self-citations, but for methods developed by the same lab that is expected and not a flaw, provided the implementations are faithful. I would cite it if I publish with SPM 25, and I would send it to review rather than desk reject. It is not a high-impact theory paper, but it is exactly the kind of artifact that peer review should cover to keep the community's main software credible.","headline":"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.","tokens_in":6097,"tokens_out":1463,"would_cite":true,"duration_ms":15104,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"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.","keywords":["statistical parametric mapping","neuroimaging software","SPM 25.01","open source","MEG/EEG analysis","OPM-MEG","Bayesian spectral decomposition","Python wrapper"],"falsifier":"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.","tokens_in":5022,"feed_emoji":"🧠","tokens_out":6492,"duration_ms":60241,"temperature":0.7,"pith_summary":"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.","feed_headline":"SPM 25.01 brings 30 years of neuroimaging analysis to Python","feed_subtitle":"First calendar-versioned release runs without MATLAB and ships new spectral and OPM tools.","key_machinery":"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.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"supporting_citations":[{"why":"Introduces the voxel-wise General Linear Model that remains the statistical core of SPM.","marker":"K. J. Friston, Holmes, et al., 1994"},{"why":"Supplies the random field theory correction for multiple comparisons that SPM uses for topological inference.","marker":"Worsley et al., 1996"},{"why":"Defines Dynamic Causal Modelling, the framework that PEB and Bayesian model reduction extend.","marker":"K. J. Friston et al., 2003"},{"why":"Establishes Voxel-Based Morphometry, a core SPM method for detecting anatomical changes.","marker":"Ashburner & Friston, 2000"},{"why":"Describes the flexible Bayesian registration method behind the Multi-Brain Toolbox.","marker":"Brudfors et al., 2020"},{"why":"Defines Bayesian Spectral Decomposition, the new spectral analysis method implemented in SPM 25.01.","marker":"Medrano et al., 2024"},{"why":"Presents the adaptive multipole models used for OPM interference cancellation in large-channel systems.","marker":"Tierney et al., 2024"},{"why":"Presents the homogeneous field correction used for OPM interference cancellation in low-channel systems.","marker":"Tierney et al., 2021"},{"why":"Supplies the FOOOF (Specparam) spectral parameterisation approach whose implementation SPM 25.01 includes alongside BSD.","marker":"Donoghue et al., 2020"},{"why":"Provides the unified beamformer framework that underpins the new M/EEG sensor fusion and BIDS support.","marker":"Westner et al., 2022"}],"fun_headline_variants":["SPM 25.01: neuroimaging analysis, Python on the horizon","SPM 25.01: open science, new methods, Python in the pipeline","SPM 25.01: 30 years strong, Python wrapper in development","SPM 25.01: 30-year-old brain suite moves toward Python"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["SPM 25.01: neuroimaging analysis, Python on the horizon","SPM 25.01: open science, new methods, Python in the pipeline","SPM 25.01: 30 years strong, Python wrapper in development","SPM 25.01: 30-year-old brain suite moves toward Python"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000579,"raw_usage":{"total_tokens":2645,"prompt_tokens":779,"completion_tokens":1866,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":395,"completion_tokens_details":{"reasoning_tokens":1779}},"tokens_in":395,"tokens_out":1866,"duration_ms":18286,"temperature":1.0,"reasoning_tokens":1779,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:31:16.185598+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[{"cited_title":"J., Marrett, S., Neelin, P., Vandal, A","cited_arxiv_id":null,"evidence_quote":"Supplies the random field theory correction for multiple comparisons that SPM uses for topological inference."},{"cited_title":"J., Harrison, L., & Penny, W","cited_arxiv_id":null,"evidence_quote":"Defines Dynamic Causal Modelling, the framework that PEB and Bayesian model reduction extend."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes Voxel-Based Morphometry, a core SPM method for detecting anatomical changes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the flexible Bayesian registration method behind the Multi-Brain Toolbox."},{"cited_title":"BSD: a Bayesian framework for parametric models of neural spectra","cited_arxiv_id":"2410.20896","evidence_quote":"Defines Bayesian Spectral Decomposition, the new spectral analysis method implemented in SPM 25.01."},{"cited_title":"M., Alexander, N., Mellor, S., Holmes, N., Seymour, R., O’Neill, G","cited_arxiv_id":null,"evidence_quote":"Presents the homogeneous field correction used for OPM interference cancellation in low-channel systems."},{"cited_title":"J., Varma, P., Sebastian, P., G ao, R., Noto, T., Lara, A","cited_arxiv_id":null,"evidence_quote":"Supplies the FOOOF (Specparam) spectral parameterisation approach whose implementation SPM 25.01 includes alongside BSD."},{"cited_title":"U., Dalal, S","cited_arxiv_id":null,"evidence_quote":"Provides the unified beamformer framework that underpins the new M/EEG sensor fusion and BIDS support."}],"review_version":1}