ds000030structural MRI / SynthSeg 2.0
Loading current state

Live research record

Updating…

Structural MRI cohort status

Loading the latest validated state.

— final outcomes
0Awaiting data265
ARCHIVED PASS—validation + checksum
EXPLICIT FAIL—retained in denominator
FINALIZED—public progress
METHODCPU × 4frozen execution lane
DATACC0public T1-weighted MRI

Cohort structure

Scanner × source ghost label

Engineering strata

Counts describe acquisition/QC strata—not biological groups. Cell imbalance limits interaction precision.

Quality control

Minimum structure-group QC score

N = —
Minimum—
Median—
Maximum—
0.00review-prioritization score1.00

Current findings

What the data support now

Claims update with verified state

Post-run technical analysis

Scanner, ghost label, and regional-volume findings

N = — · HC3
QC median—minimum structure-group score
Low-fence count—distributional flag, not failure
Regional FDR—q < 0.05 across 3 contrasts
Analysis hash—method + inputs + software

Minimum-QC factorial model

Absolute QC-score differences from the reference cell: scanner 35343.0 with no ghost label.

TermEstimate95% CIFDR q

Observed QC by acquisition cell

Descriptive values; the smallest scanner×ghost cell limits interaction precision.

Scanner / labelNMedianIQR

Intracranial-normalized regional models

One hundred exploratory regions; Benjamini–Hochberg correction applied separately within each contrast.

Lowest q-value per contrast

Effect estimates are technical scanner/label associations from log(volume ÷ total intracranial volume) models. They do not establish anatomical accuracy, biological effects, or clinical differences.

Visual quality control

Representative review summary

Aggregate counts only
Reviewed—
Pass—
Flagged—

Review scope

    No participant identifiers or subject-level MRI images are published on this public surface. Representative-slice review does not establish complete voxelwise accuracy.

    Pipeline

    Processing and archive path

    Structural T1w only
    1. 01
      Discover + bind

      Resolve 265 local git-annex T1w files; validate 3D geometry; freeze SHA-256 manifest.

    2. 02
      Run standard SynthSeg

      Pinned FreeSurfer image, cortical parcellation, regional volumes, QC, and a verified L4 TensorFlow GPU.

    3. 03
      Validate independently

      Discrete labels, pinned CSV schemas, finite values, source geometry, configuration binding.

    4. 04
      Archive in eights

      Copy to hash-versioned Drive namespace, checksum-check, then release local staging.

    Execution record

    Measured invocations

    Reproducibility

    Frozen provenance

    Configuration-bound
    Automated validation contract

      Interpretation boundary

      Research engineering, not diagnosis

        Primary sources

        References