Featured method · doctoral research
The surface changes with age. The hard part is asking it the right question.
A 3D surface contains change at many geometric scales. My doctoral work showed how to isolate the age-related part of that change—and why a method should stop where the evidence stops.
The inversion problem
Age causes remodeling. A surface does not uniquely reveal age.
Conventional phase systems turn a continuous, variable surface into an ordered label and then work backwards to age. Morphometric prediction has the same risk in a more technical form: it can make an inverse relationship look precise even when the biological effect is non-linear, overlapping and weak at older ages.
The productive move was to constrain both sides of the problem: focus on the age range where directional remodeling remains visible, then retain only the geometric scales that actually carry that signal.
Interactive method
From a scanned surface to a defensible boundary
The animation is schematic: it explains the analytical logic without publishing a specimen or research dataset.
01 / 05
Capture the surface
High-resolution surface scans preserve the fine relief of the pubic symphysis. Acquisition quality matters because roughness is the signal being tested, not noise to erase casually.
02 / 05
Build correspondence
Fixed landmarks and surface semilandmarks put homologous locations into a shared coordinate system, allowing local deformation to be compared across surfaces.
03 / 05
Decompose the deformation
Thin-plate-spline bending energy separates deformation into partial warps: components ordered from broad, low-frequency form to increasingly fine surface detail.
04 / 05
Keep the informative scales
For younger adults, the useful age-related signal concentrated in the broadest scales. A low-pass filter retained the first partial warps and rejected finer variation that did not improve the relationship.
05 / 05
Stop at the evidence boundary
The association weakened around age 50. The defensible result is therefore a constrained method for younger adults—not a universal age estimate from one surface.
Evidence from the study
A useful signal, under explicit conditions
The figures below describe the published workflow; they are not a clinical or forensic calculator.
- 440
- surface-scanned pubic bones in the full study sample
- 102
- fixed and surface semilandmarks used to represent each surface
- 7
- broadest partial-warp amplitudes retained at the strongest filtered relationship
- −0.648
- peak reported correlation in the selected right-side younger-adult subset
- < 50
- the essential age boundary for the reported relationship
What the work refuses to claim
A boundary is a result, not a failure.
- 01
One pubic-symphysis surface does not support a universal adult age estimator.
- 02
Fine geometric detail is not automatically useful signal; filtering must be validated.
- 03
After roughly age 50, the tested relationship is too weak for the same inverse-regression claim.
- 04
Older-age estimation needs additional, asynchronous indicators rather than greater confidence in one surface.
Read the method
Three connected outputs
- 2021 · Journal article
Correlation of the human pubic symphysis surface with age-at-death
The bandpass-filter workflow and its quantitative boundary.
Open DOI - 2021 · Journal article
The outline is sexually dimorphic and changes with age
A complementary analysis showing that outline trajectories vary by sex and lose age signal around the same boundary.
Open DOI - 2021 · Book chapter
The fallacy of forensic age estimation from morphometric quantifications
The inverse-regression problem, why broad-range estimates fail and what a multi-indicator solution requires.
Open DOI