Multiparametric CT-Based Adult Age Estimation Using Thoracic Vertebral Degenerative Changes, Vertebral Bone and Paraspinal Muscle Attenuation: A Regression-Based Approach.

Background/

Objectives

Adult age estimation remains challenging because degenerative processes lack uniform progression and show substantial interindividual variability. This study aimed to evaluate the association between thoracic vertebral degeneration and chronological age using multidetector computed tomography (MDCT) and to develop regression-based age estimation models.

Methods

A total of 240 individuals (120 females and 120 males) were included. Four parameters-osteophyte formation, facet joint degeneration, vertebral corpus bone attenuation, and paraspinal muscle attenuation-were assessed, generating 15 derived variables used for regression models. Univariable and multivariable linear regression analyses were conducted in combined and sex-stratified samples. Internal validation was conducted using 1000 bootstrap resamples.

Results

Among single variables, linear osteophyte measurement (L0) showed the strongest association with age in the combined sample (R2 = 0.682), with a root mean squared error under 10 years. Two multivariable regression models were constructed. In the combined development sample, model 1, which allowed both morphological and attenuation-based variables to enter the selection procedure, achieved an R2 of 0.84 with a root mean squared error of 6.93 years. Morphology-based model 2 yielded an R2 of 0.75 and a root mean squared error of 8.58 years. Bootstrap internal validation showed modestly lower performance than the apparent estimates.

Conclusions

Thoracic vertebral degenerative and attenuation-based parameters assessed by MDCT demonstrate potential as a complementary method for adult age estimation. The proposed dual-model strategy may offer flexibility for evaluating thoracic vertebral parameters under different forensic conditions. External validation in multicenter and postmortem cohorts is needed to establish the generalizability of the models and assess their suitability for routine forensic application.