Cervical Spinal Cord Morphometry Charts

Spinal cord morphometry provides essential biomarkers of neurological health, but clinical interpretations are confounded by inter-subject variability and a lack of normative references across the full human lifespan. We address this gap by generating the first comprehensive lifespan charts for cervical spinal cord morphometry. We leveraged 30 population-based brain MRI datasets, aggregating 78,269 scans from 41,042 individuals (ages 0–100) whose imaging protocols included cervical cord coverage. To overcome contrast variability, we employed a state-of-the-art contrast-agnostic deep-learning segmentation method, extracting cross-sectional area (CSA), anteroposterior (AP) and right–left (RL/transverse) diameters, and shape indices (compression ratio, eccentricity, and solidity) from C1 to C7. Normative trajectories were modeled using Generalized Additive Models for Location, Scale, and Shape (GAMLSS). The resulting charts reveal distinct nonlinear lifespan changes: rapid growth through childhood and adolescence, peak maturation in early-to-mid adulthood (e.g., the mid-30s for CSA), followed by gradual decreases. Significant regional variations along the cervical cord and consistent sex differences (males > females for size metrics) were quantified. Spinal cord trajectories showed strong temporal coupling with brain white-matter and brainstem volumes, suggesting integrated central nervous system development and aging. These lifespan charts provide a normative framework for age- and sex-specific centile scoring of individual spinal cord morphometry. This release provides the centile curves, RDS chart files, software, templates, and worked examples needed to score and align new datasets. What to Download Download the single file spinal-cord-lifespan-charts-complete.zip. It contains the current centile tables, RDS chart files, alignment software, input templates, and worked examples in the correct directory structure. No additional model or centile download is required. After unzipping, the shortest cross-platform route is: Rscript install_dependencies.RRscript align_to_charts.R examples/inputs/BMS.csv results/BMS The first command installs the required R packages and normally needs to be run only once. Centile CSV files The reference/centiles/ folder contains one CSV file for each charted spinal cord, brain, or brainstem feature. For example, reference/centiles/mean_area_L3_centiles.csv gives female and male C3 cross-sectional area values for the 1st–99th centiles across age. These files can be opened directly for reference, plotting, or reporting. These comprise 49 spinal cord charts - cross-sectional area, anteroposterior and right–left diameters, compression ratio, eccentricity, solidity, and length at C1–C7 - and nine FreeSurfer-derived brain or brainstem volume charts: Brain-Stem, BrainSegVol, BrainSegVolNotVent, CerebralWhiteMatterVol, CortexVol, eTIV, SubCortGrayVol, TotalGrayVol, and TotalThalamusVol. GAMLSS model RDS files The reference/parameters/ folder contains one RDS file for each charted feature, for example reference/parameters/mean_area_L3_reference.rds. These files contain the age- and sex-specific GAMLSS chart parameters used by the software to calculate centiles, align new datasets, and return harmonized feature values. Aligning a new dataset to the charts The software accepts a wide CSV with one row per participant or visit, subject, age, sex_label (female or male) or sex (0 or 1), a group or status column with controls labeled CN, and any subset of supported feature columns (i.e., spinal cord or brain morphometry). A dataset column is optional and should be used when the file contains multiple sites or acquisition batches. For each dataset and feature, the software uses the controls to estimate location and scale offsets, applies the same offsets to all participants, and calculates raw and aligned centiles and harmonized feature values. To align your own input: Rscript align_to_charts.R INPUT.csv OUTPUT_DIRECTORY Input template and outputs A ready-to-use multi-subject example is included at examples/alignment_input_template.csv. The same table is also provided as examples/spinal_cord_chart_alignment_input_template.xlsx, with a second sheet listing all supported feature names. Use any subset of the feature columns available in your dataset. chart_scores.csv is the main output. It keeps the original input columns and adds raw centiles, aligned centiles, and harmonized values. aligned_centiles.csv and aligned_features.csv provide smaller result tables, and qa_pdf/ contains plots for checking the alignment. Worked examples The archive includes four worked examples: BMS (controls and participants with multiple sclerosis), DLBS, MASiVar, and UCLA_LA5C. Each has an input CSV in examples/inputs/ and completed outputs in examples/outputs/. BMS is the smallest example and the easiest place to start. Together, the examples show the same workflow with different combinations of spinal cord and brain/brainstem features. To regenerate all four example outputs: examples/run_all_public_examples.sh Producing spinal cord features with SCT For researchers starting from T1-weighted images, spinal cord features can be produced with the Spinal Cord Toolbox. In our workflow, each image was processed with: sct_deepseg spinalcord -i T1w.nii.gz -qc qcsct_deepseg totalspineseg -i T1w.nii.gz -step1-only 1 -qc qcsct_process_segmentation \ -i T1w_seg.nii.gz \ -discfile T1w_step1_levels.nii.gz \ -vert 1:9 \ -perlevel 1 \ -o CSA.csv The resulting CSA.csv contains one row per vertebral level and measurements including area, anteroposterior and right–left diameters, eccentricity, solidity, and length. Segmentations and vertebral labels should be reviewed visually before the measurements are used. After QA/QC, multiple CSA.csv files can be converted into an alignment-ready wide table: python3 -m pip install -r requirements_python.txtpython3 scripts/prepare_sct_csa_for_alignment.py \ --manifest=examples/sct_csa_manifest_template.csv \ --out-csv=outputs/my_dataset_alignment_input.csv The manifest contains one row per subject/session and, at minimum, csa_csv, subject, age, sex_label or sex, and preferably group with controls labeled CN. Session is optional. Notes for users QA/QC filter all input measurements before alignment and confirm that feature names and units match the supplied references. Use female or male in sex_label; numeric sex values 0 and 1 are also accepted. Use CN to identify the controls used to estimate alignment. If no group/status column is supplied, the software warns and treats all rows as controls; use this option only for a known control-only file. By default, one eligible observation per unique control is used. The minimum of 10 unique controls is an operational floor; at least 100 controls are recommended, and smaller samples require careful review of diagnostics and sensitivity analyses. Missing chart-backed features are ignored. Observations outside a feature’s reference age range are retained in the output but are not scored for that feature. Review the generated QA plots, convergence information, and alignment diagnostics before interpreting aligned centiles. Centiles are normative positions, not diagnostic cutoffs.