MRI Tractography Thresholds Reshape Growth Plate Signals in Teenagers

Diffusion tensor imaging of the growth plate has emerged as one of the most intriguing noninvasive windows into skeletal development, but a new study suggests that one seemingly technical choice—the threshold used to reconstruct tracts—can dramatically change what the images appear to say about a teenager’s growth. In a longitudinal analysis of 45 healthy adolescents aged 12 to 17, researchers found that correlations between tractography-derived metrics of the distal femoral physeal-metaphyseal complex and annualized height velocity were highly sensitive to the fractional anisotropy, or FA, thresholds applied during tract reconstruction, while two other metrics remained remarkably stable across every setting tested.

The work, published in European Radiology Experimental, was a methodological sub-analysis of a prospective cohort study conducted at two tertiary pediatric centers in the northeastern United States. Participants were recruited from general pediatric practices between July 2020 and July 2022, and each was imaged twice at a one-year interval. Height and weight were measured by trained research staff at baseline and follow-up using standardized equipment, and height velocity was defined as the annualized change in height over the 12-month period. The authors emphasize that the analysis was descriptive and methodological in nature, with p-values reported for descriptive purposes and no adjustment for multiple comparisons applied.

The conventional FA thresholds used in prior growth plate tractography studies are a minimum of 0.15 and a maximum of 1.0. These values were originally developed for highly organized tissues and, according to the researchers, may systematically exclude the less organized hypertrophic and transitional zones of the growth plate—the very regions thought to drive longitudinal bone growth. To test whether lowering the thresholds would better capture growth-sensitive features, the team evaluated nine previously untested threshold combinations, pairing minimum FA values of 0.05, 0.10, and 0.15 with maximum FA values of 0.40, 0.50, and 0.60, alongside the conventional 0.15 to 1.0 setting.

Imaging was performed at 3 Tesla using a fat-suppressed single-shot spin-echo echo-planar sequence with 20 non-collinear diffusion directions, b-values of 0 and 600 sec/mm2, and a voxel size of 12 cubic millimeters. One center used a GE Signa Premier scanner with an 18-channel knee coil, while the other used a Siemens Magnetom Prisma with a 15-channel coil. Key parameters affecting diffusion quantification, including voxel size, b-values, and diffusion direction scheme, were harmonized across sites, although repetition and echo times were optimized independently for each platform. A prior comparison of the two platforms using Bland-Altman plots and intraclass correlation had shown high reproducibility.

Physeal regions were manually segmented on the baseline images by a postdoctoral researcher with more than one year of segmentation experience and verified by two senior pediatric radiologists with 7 and 14 years of DTI research experience. For this analysis, only the distal femoral physis was evaluated. Deterministic tractography was then performed using a Fiber Assignment by Continuous Tracking algorithm, and for each threshold setting the team extracted tract volume, defined as the total volume of voxels transected by reconstructed tracts, along with tract-weighted FA and apparent diffusion coefficient, or ADC, computed as mean values within those voxels.

The headline result concerned tract ADC. Partial Pearson correlations controlling for Tanner stage showed that tract ADC of the distal femoral physeal-metaphyseal complex was strongly and positively associated with height velocity across all ten threshold combinations, with correlation coefficients ranging from 0.55 to 0.63 and nominal p-values below 0.001. The strongest association appeared at a threshold of 0.05 to 0.60, where r reached 0.63 with a 95 percent confidence interval of 0.42 to 0.78. Changing the minimum threshold from 0.05 to 0.15 produced only modest attenuation of roughly 0.05 in the correlation, and varying the maximum threshold from 0.40 to 1.00 had minimal effect. By Cohen’s conventions, these are large effect sizes.

Tract FA told a similarly stable story, but in the opposite direction. Across every threshold combination, tract FA showed consistent negative partial correlations with height velocity, ranging from -0.38 to -0.45. The largest association was observed at a threshold of 0.15 to 0.60, where r equaled -0.45 with a 95 percent confidence interval of -0.66 to -0.18 and a nominal p-value of 0.002. In other words, adolescents growing faster tended to have lower directional organization of water diffusion in their growth plate tissue, a pattern the authors interpret as reflecting a less restricted microstructural environment during rapid growth.

Tract volume was a different matter entirely. At the most permissive minimum threshold of 0.05, tract volume showed nominally significant positive correlations with height velocity: 0.37 at a threshold of 0.05 to 0.40, 0.34 at 0.05 to 0.50, and 0.33 at 0.05 to 0.60. When the minimum threshold was raised to 0.10, the correlations attenuated to between 0.10 and 0.20 and lost nominal significance. At the restrictive minimum of 0.15, the coefficients shifted toward negative values without reaching significance, including the conventional threshold of 0.15 to 1.00, which yielded r of -0.26 with a confidence interval spanning zero. A representative early-pubertal subject with a height velocity of 8.3 centimeters per year illustrated the magnitude of the difference: tract volume was 25.5 cubic millimeters at the permissive 0.05 to 0.40 threshold but only 6.1 cubic millimeters at the conventional setting.

The authors argue that the pattern makes biological sense. Lower minimum thresholds appear to capture the less organized hypertrophic and transitional zones linked to longitudinal growth, which conventional thresholds systematically underestimate, while restricting the maximum FA value helps exclude highly organized, mineralized tissue that becomes dominant as the growth plate matures and that may not reflect active growth. They point to prior animal work showing that physeal FA rises toward skeletal maturation as mineralized matrix replaces actively dividing chondrocyte columns, and to evidence that elevated ADC reflects diffusion within the transitional zone characterized by ossified matrix interspersed with hypertrophic chondrocyte columns. They also acknowledge the counterargument that permissive thresholds risk including isotropic noise, but note that the significant tract volume correlation at the 0.05 minimum threshold, which vanished at higher thresholds, suggests the additional volume represents a meaningful biological signal rather than stochastic artifact.

The team proposes minimum and maximum FA values of 0.05 and 0.40 as starting points for optimization in future studies of growth plate tract volume, while stressing that these are not definitive clinical standards. The study’s limitations are substantial: the sample of 45 was not powered to estimate population-level effect sizes, the age range of 12 to 17 years means the findings may not generalize to younger children or cohorts with decelerated growth, and residual cross-platform bias between the two scanner vendors cannot be entirely excluded in the absence of a traveling-subject calibration study. The authors recommend that investigators report results across multiple threshold combinations and validate findings in larger cohorts specifically powered to estimate the underlying correlations before any single threshold is adopted. Within those limits, the message is clear: tract ADC and FA are robust markers of growth plate activity that survive any reasonable threshold choice, but tract volume—the metric most intuitively tied to the amount of growth-active tissue—depends critically on how the reconstruction is tuned, underscoring the need for standardized tractography parameters in pediatric skeletal growth imaging.

Subject of Research: How fractional anisotropy threshold selection affects diffusion tensor imaging tractography correlations with adolescent height velocity in the growth plate

Article Title: Sensitivity of physeal-metaphyseal tractography correlations with height velocity for fractional anisotropy threshold selection

Article References: Perdomo-Luna, C., Jaramillo, D., Santos, L., Mostoufi-Moab, S., Galvis, I., Do, H., Jambawalikar, S., Kammen, B., & Kvist, O. (2026). Sensitivity of physeal-metaphyseal tractography correlations with height velocity for fractional anisotropy threshold selection. European Radiology Experimental, 10(1), Article 161. https://doi.org/10.1186/s41747-026-00826-8

Image Credits: AI Generated

DOI: 10.1186/s41747-026-00826-8

Keywords: diffusion tensor imaging, tractography, growth plate, physeal-metaphyseal complex, fractional anisotropy, height velocity, adolescents, MRI, skeletal maturation, pediatric imaging, distal femoral physis, apparent diffusion coefficient