AI Development September 7, 2026 ~13 min SPM 25.01.02 SPM12

SPM 25.01.02 or SPM12: 2026 Neuroimaging Project Choice

This guide helps neuroimaging researchers and university support teams choose between SPM 25.01.02 and SPM12. It separates new projects, active papers, legacy scripts, Standalone use, and Apple Silicon deployment, with a dual-track migration method and acceptance criteria.

As of September 7, 2026, the official SPM releases page marks SPM 25.01.02 as the latest stable release, while the official SPM12 page describes SPM12 as an older, unsupported version. For a new neuroimaging project, choose SPM 25.01.02. For an active paper, historical replication, or script-heavy project, keep SPM12 and validate both versions before migrating.

This article is for: graduate students starting fMRI, PET, EEG, or MEG analysis; researchers completing a paper or reproducing an older result; and university technical staff managing MATLAB, Apple Silicon, and shared research environments.

The decision starts with project status, not the newest version

The version question is not simply “which release has more features?” It is a risk decision. A new project has no result history to protect, so a currently supported stable release is the cleaner baseline. An active project has analysis decisions, generated files, scripts, and publication claims that may depend on the old environment.

The official versioning documentation explains how SPM identifies releases and development versions. That distinction matters because a pre-release build should not be treated as a production baseline merely because it has a newer-looking version number. The 26.01.alpha21 build is outside the recommendation in this article because it is a pre-release, not a stable production target.

Use this default rule:

New project or first SPM deployment: start with SPM 25.01.02.
Paper in analysis, submission, or review: retain SPM12 until the key outputs are frozen.
Historical replication: reproduce with the original SPM12 environment first.
Custom MATLAB scripts or third-party extensions: use a dual-track regression.
No MATLAB license: assess the Standalone route with a representative task before committing.
Apple Silicon deployment: isolate each SPM version instead of sharing paths and writable folders.

The score below is a decision aid, not a performance benchmark. A higher score means lower migration risk for that audience.

Research situation	SPM 25.01.02	SPM12	Recommended operating mode
New fMRI, PET, EEG, or MEG project	5/5	2/5	Start with SPM 25.01.02
Active paper using established results	2/5	5/5	Keep SPM12; test the new version separately
Historical replication	2/5	5/5	Reproduce with SPM12 first
Large custom script library	3/5	4/5	Run both until regression passes
Third-party toolbox dependency	3/5	4/5	Check each extension and error log
MATLAB unavailable	3/5	3/5	Compare Standalone suitability with a pilot
Apple Silicon transition	4/5	3/5	Use isolated environments and architecture checks

New projects should establish a supported baseline

First step: define what “working” means

A new project should not begin with an old lab folder copied from a previous student. That folder may contain hidden MATLAB paths, startup commands, compiled MEX files, default settings, or assumptions about input filenames.

For SPM 25.01.02, define a small acceptance package before processing the full dataset:

Record the SPM version output and MATLAB release.
Launch the graphical interface successfully.
Run one minimal batch job with known input data.
Confirm that expected result files are created.
Save the batch script, settings, and environment notes.
Repeat the same test after a clean reinstall or on a second isolated node.

The official SPM installation instructions should be the installation baseline. They are more reliable than copying a lab-specific path setup whose original dependencies may no longer exist.

The point is not to prove that every toolbox will work. It is to prove that the core workflow is visible, executable, and reproducible before the project accumulates dozens of subjects.

Why the current stable release is the better default

SPM 25.01.02 gives a new project a documented release target that can be recorded in a methods section and reproduced by another member of the lab. Starting with SPM12 may appear safer because many courses and older protocols use it, but that safety is often familiarity rather than support status.

A new project can still use a legacy extension if the research design requires it. The correct approach is to test that extension inside the new baseline, not to make an unsupported old version the default for every future analysis.

Research note: “The interface opened” is not an acceptance test. A valid baseline also needs a completed batch, expected output files, and a recorded version path.

Active papers should protect their existing result chain

Why upgrading during submission is a poor trade

When a manuscript is being analyzed, submitted, or reviewed, changing SPM versions creates a new variable at the worst possible point. Even if the final statistical interpretation remains similar, the team may need to explain differences in preprocessing, model estimation, contrasts, images, or quality-control outputs.

For an active SPM12 project, preserve:

The original SPM12 directory or environment image.
The exact MATLAB release and operating system.
The SPM.mat files.
Batch scripts and startup files.
Default settings used by the lab.
Input data references and preprocessing order.
Extension versions and any compiled MEX files.
Representative first-level and group-level outputs.

The official SPM12 compatibility information identifies its support status. That status does not mean an existing study must be abandoned. It means the team should understand that preserving a known environment and starting a new supported project are different management decisions.

In a replication study, reproduce before modernizing

A historical replication should first answer: “Can the published or archived workflow reproduce its reference output?” That answer requires the original environment, or the closest documented substitute, before testing SPM 25.01.02.

Use a read-only copy of the original data and scripts. Work from a separate project directory. Compare predetermined outputs such as subject counts, design dimensions, mask sizes, parameter estimates, contrast images, and final tables. The exact criteria depend on the study, but they must be defined before inspecting the new result.

If the old result cannot be reproduced, migrating to a new SPM release will make diagnosis harder. First determine whether the discrepancy comes from missing data, changed input ordering, MATLAB behavior, a path problem, or the SPM environment itself.

Custom scripts need regression evidence, not a successful launch

SPM12 batch scripts and the new release

An SPM12 batch script may run under SPM 25.01.02, but that cannot be assumed from a successful GUI launch or from the script starting without a syntax error. A batch can complete while producing different fields, defaults, paths, or intermediate files.

Review each script for:

Function names and changed call signatures.
Batch field names and nested structures.
Default parameter assumptions.
Relative and absolute path handling.
Use of global variables or startup files.
Calls to external toolboxes.
MEX binaries compiled for a different architecture.
File naming and output overwrite behavior.

The SPM compilation documentation explains why compiled components need separate attention. The MathWorks MEX compatibility guide is also relevant when a script depends on native binaries. A MATLAB script can be portable while its MEX dependency is not.

Second step: build a dual-track regression

Keep SPM12 unchanged as the control environment. Create a separate SPM 25.01.02 environment with its own MATLAB path, startup file, SPM directory, temporary folder, and output directory.

Then run the same anonymized representative task in both environments:

Use identical input data and subject order.
Use copies of the original scripts.
Capture the command-window log.
Compare intermediate and final output files.
Record warnings, missing fields, and runtime errors.
Investigate every unexplained result difference.
Set a written stop condition before testing the full dataset.

A pass does not require byte-for-byte identity for every file. It does require a scientifically justified equivalence standard. For example, the lab may require matching subject inclusion, design structure, quality-control outcomes, and prespecified statistical outputs within an agreed tolerance.

The SPM12 release notes can help explain historical behavior, but release notes are not proof that a third-party extension remains compatible. For extensions, use the project’s own documentation and the actual error log.

MATLAB access changes the route, not the acceptance standard

Standalone can help, but it is not a universal substitute

Researchers without a MATLAB license should evaluate SPM Standalone against the actual project. The official Standalone documentation describes the deployment model and its dependencies.

Check these items before choosing it:

Whether the required graphical interface is available.
Whether the planned batch workflow runs correctly.
Whether the needed MATLAB Runtime is installed and supported.
Whether custom MATLAB code can run.
Whether the project depends on additional MATLAB toolboxes.
Whether required third-party SPM extensions support Standalone.
Whether outputs can be generated and transferred in the intended environment.

Standalone may fit a workflow that uses core SPM functions and fixed batch files. It is a weak assumption for a project built around custom MATLAB functions, interactive debugging, or extra toolboxes.

Third step: run a small license decision test

Use a de-identified sample rather than the entire study. Run one representative preprocessing task, one model specification, and one output export. If the project depends on a custom function, include it in this test.

Do not choose Standalone only because it appears cheaper or avoids a campus license request. If the first real analysis fails after the project has been designed around that route, the saved license cost will not compensate for rework.

Apple Silicon requires separate environments and explicit checks

Apple Silicon changes the deployment details, but it does not change the scientific decision rule. A Mac can host both versions, yet a shared MATLAB path or a shared writable SPM directory can make the result difficult to interpret.

MathWorks documents Apple Silicon support and platform requirements. Use that documentation to confirm the MATLAB release and architecture policy before selecting the host environment.

Keeping both versions on one Apple Silicon Mac

On an Apple Silicon Mac, separate at least these items:

SPM 25.01.02 and SPM12 directories.
MATLAB path definitions.
Startup files.
Temporary and writable output folders.
MEX binaries and compiled dependencies.
Batch scripts and configuration files.
Logs and result directories.

Avoid a single startup file that silently adds both SPM versions. A script that works only because the wrong version appears earlier on the path is not a reproducible setup.

Fourth step: validate architecture and remote behavior separately

Run these checks for each version:

Confirm which MATLAB and SPM directories are active.
Check whether required MEX files load on Apple Silicon.
Start the GUI through the intended local or remote method.
Run a short batch without disconnecting.
Reconnect and verify that the process completed.
Download the output and compare it with the local control result.
Record the host architecture, MATLAB release, SPM version, and path state.

Remote screen latency is an interaction problem, not evidence of slow SPM computation. Conversely, a responsive VNC session does not prove that a long batch will survive a dropped connection. Treat interface behavior, process continuity, and scientific output as three separate tests.

Validation area	Pass condition	Stop and investigate when
Version identity	Log clearly shows the intended SPM release	The active path contains both versions
GUI launch	Interface opens and loads the intended modules	Startup fails or opens an unexpected path
Batch execution	Representative job completes without unexplained errors	A job stops after disconnect or changes outputs
Native components	Required MEX files load on the host architecture	A binary is missing or built for another platform
Output integrity	Expected files and metadata are present	Files are missing, overwritten, or inconsistent
Reproducibility	Control and candidate results meet preset criteria	Differences cannot be explained

Fifth step: give the lab a lifecycle rule

A lab manager should not ask every project to migrate on the same date. Assign a version policy by project lifecycle:

New projects: SPM 25.01.02 by default.
Frozen or active studies: preserve SPM12 with documented access.
Migration candidates: run both versions on a representative task.
Custom-toolbox projects: require extension-by-extension sign-off.
Projects without MATLAB: approve Standalone only after a pilot.
Long-running remote workflows: require reconnect, logging, and output-transfer tests.

Every project record should include SPM version, MATLAB release, macOS version, architecture, extension list, script revision, input-data identifier, output location, and rollback environment. The rollback environment should be usable by another team member, not stored only on the original researcher’s laptop.

A migration should stop when a key output changes without a documented scientific explanation, when a required extension fails, when native code cannot load, or when the team cannot reproduce the control result. “We can troubleshoot it later” is not a migration criterion.

A remote Mac can be a controlled validation node

If the lab has Linux and Windows systems but no Mac suitable for isolated testing, a remote Apple Silicon Mac can serve as a temporary validation node. This is most useful when the team needs a real macOS environment, root access for setup, and a clean place to keep SPM12 and SPM 25.01.02 apart.

VNCMac provides remote Mac access options that can be considered for a short validation cycle rather than treated as a replacement for every research workstation. The relevant checks are not only whether the desktop is reachable, but whether the environment supports the required MATLAB route, batch continuity, file transfer, and secure handling of de-identified data.

For regional access planning, the available remote Mac locations can help the technical team choose a connection path that fits its users. Keep the data boundary explicit: sensitive or restricted research data may require institutional approval before being copied to any externally hosted machine.

A sensible sequence is:

Prepare de-identified input data.
Install or provision the two isolated SPM environments.
Record MATLAB and macOS details.
Run the same representative script in both versions.
Save logs and output checksums or file inventories.
Review differences with the project scientist.
Decide whether to keep SPM12, migrate, or operate both tracks.

Final decision for each audience

For a student starting a new fMRI, PET, EEG, or MEG project, SPM 25.01.02 is the default choice. Build a clean baseline and document it before importing lab history.

For a researcher preparing a paper or replication, SPM12 remains the safer control environment. Do not replace it during a critical publication milestone. Test SPM 25.01.02 on a copy and migrate only after the predefined outputs pass.

For a script or toolbox maintainer, neither version is automatically safe. The correct answer is dual-track regression, with every unexplained difference treated as a release blocker.

For a student without MATLAB, Standalone may be appropriate, but only when the required GUI, Runtime, batch, custom code, and extensions pass a representative pilot.

For a university support team, the strongest policy is not “everyone upgrades.” It is a lifecycle policy: supported stable release for new work, preserved legacy environment for active work, and documented dual-track validation for migration.

If the current setup is only a Linux or Windows machine, it has three practical weaknesses for this decision: it cannot provide a native macOS validation path, it may force researchers to mix an old SPM environment with unrelated lab dependencies, and it does not by itself test Apple Silicon behavior or remote GUI continuity. Buying a Mac for a single migration check can also leave the hardware underused after the project decision is complete.

For a short validation window, renting a remote Apple Silicon Mac through VNCMac can be the more controlled option: keep SPM12 and SPM 25.01.02 isolated, run one de-identified representative task, and make the version decision from evidence rather than from interface familiarity. Long-term heavy workloads, restricted data, or instrument-connected experiments may still justify institution-owned hardware; temporary compatibility work does not always require a new purchase.