MRI Post-Processing and Morphometric Analysis in Epilepsy

MRI post-processing uses computational methods to analyse structural brain images for abnormalities that may be too subtle to recognise on routine visual inspection. In epilepsy, its main role is finding possible epileptogenic lesions—particularly focal cortical dysplasia—in people whose conventional MRI has been reported as normal or equivocal.

Why “MRI-negative” may not really mean lesion-negative

Some epileptogenic lesions are extremely subtle. They may cause only slight cortical thickening, abnormal gyration, blurring of the grey–white junction or changes in signal intensity.

A 2025 systematic review of MRI-negative epilepsy found that post-processing substantially increased diagnostic yield, reinforcing that a negative visual read does not always mean no structural lesion exists.

What morphometric analysis does

Morphometric techniques measure features of brain structure mathematically rather than relying entirely on visual inspection.

They can compare cortical thickness, grey–white contrast, tissue intensity, sulcal shape and other structural features with expected patterns.

Voxel-based methods

Voxel-based approaches examine small three-dimensional units of the MRI and look for statistically unusual tissue characteristics.

The Morphometric Analysis Program, including modern versions such as MAP18, is one example used in focal cortical dysplasia research and clinical practice.

Surface-based methods

Surface-based analysis reconstructs the cortical surface and can measure features such as cortical thickness, sulcal depth and grey–white boundary characteristics.

These methods can be useful when abnormalities follow the shape of the cortex rather than forming a clearly defined lesion.

Focal cortical dysplasia

Focal cortical dysplasia is the main clinical target for many MRI post-processing tools.

Type II lesions can sometimes be visually apparent, while type I and other subtle dysplasias may be much harder to recognise.

Machine learning and artificial intelligence

More recent systems use machine-learning or deep-learning methods to recognise combinations of imaging features associated with cortical dysplasia.

Examples under active study include surface-based classifiers and multicentre tools such as MELD-related approaches.

Post-processing does not make the diagnosis by itself

A highlighted region is a candidate abnormality, not proof of the seizure-onset zone.

False positives occur, and every candidate must be reviewed on the original MRI and compared with seizure semiology, EEG, PET, MEG, source imaging and other clinical information.

Image quality matters

Post-processing cannot reliably rescue poor source imaging.

Motion, thick slices, low spatial resolution or inappropriate MRI sequences can produce misleading results. A 2024 study of MAP18 emphasised screening MRI quality before relying on morphometric analysis.

Why expert re-review is essential

Computational maps often work best as a guide for an experienced neuroradiologist.

When software flags a region, the radiologist can return to the original T1, FLAIR and T2 images and look for subtle corresponding abnormalities.

Does a positive map change surgery?

It can. A credible abnormality may influence where PET images are re-examined, where SEEG electrodes are placed or which cortical region is considered for resection.

Historical MAP studies have also found that complete resection of concordant MAP-positive abnormalities can be associated with better seizure outcomes.

What recent evidence shows

A 2025 review of voxel- and surface-based morphometry found considerable promise but also substantial variation between techniques, study designs and reference standards.

This means post-processing is clinically useful, but performance figures from one laboratory cannot automatically be applied to every scanner, population or software package.

Children

Paediatric analysis is more complex because cortical thickness, myelination and tissue contrast change during development.

Age-appropriate reference data and high-quality paediatric protocols are therefore important.

7T MRI and post-processing

Ultra-high-field MRI and computational post-processing are complementary rather than competing techniques.

Higher field strength may reveal additional anatomical detail, while post-processing may expose subtle statistical abnormalities even on high-quality 3T imaging.

Worldwide access

The main barriers are specialised software, scanner-compatible reference datasets, technical expertise and experienced neuroradiological review.

Open and multicentre tools may eventually make advanced analysis more widely accessible, but clinical validation remains essential.

The central point

MRI post-processing is a lesion-detection aid, not an automated diagnosis. Its greatest value is helping specialists find subtle abnormalities in focal epilepsy that conventional visual review may have missed.

Sources

Evidence checked: September 2026.