The research, published in the International Journal of Legal Medicine, was led by Sushil Kumar Battan and colleagues at the Postgraduate Institute of Medical Education and Research in Chandigarh together with collaborators at Chandigarh University. The team carried out a retrospective cross-sectional analysis of computed tomography scans from 1,972 individuals aged between 18 and 80 years, all drawn from a northwestern Indian population. Rather than relying on the mean tissue depths that have dominated facial reconstruction practice for decades, the researchers measured both total craniofacial soft tissue thickness and the subcutaneous fat layer thickness separately, at 73 standardized craniofacial landmarks.

The distinction matters because the total thickness at any point on the face is not a single uniform material. It is a composite of skin, muscle, fascia, glands and fat, and only the subcutaneous fat component changes dramatically with age, sex, nutrition and body composition. Most existing reference tables, including large in vivo databases assembled for European populations, collapse all of these layers into a single number. By separating the adipose from the non-adipose contribution, the Indian dataset offers reconstructers a two-parameter description of facial architecture, which in principle allows far more nuanced modeling of how a face would have looked in life.

Technically, the measurements were extracted from CT imaging, which provides the millimeter-scale resolution needed to delineate the boundary between the bone, the deeper soft tissues and the subcutaneous fat plane. The statistical analysis combined classical tools with machine learning. Descriptive statistics established the baseline distributions at each landmark, independent t-tests quantified differences between the sexes, analysis of variance tested variation across age groups, and Pearson correlation assessed how tightly the two thickness measures track each other. On top of this, the team trained regression and machine learning algorithms to predict tissue thicknesses and to evaluate how well different variables explain the observed variation.

The results confirm that total soft tissue thickness consistently exceeds fat layer thickness at every one of the 73 landmarks, underscoring that a substantial portion of facial volume comes from muscle and other non-adipose structures. At the level of individual landmarks, the correlation between the two measures was strong, with a coefficient of determination of roughly 0.785, meaning that about three quarters of the landmark-level variation in total thickness can be accounted for by the fat layer alone. Interestingly, that relationship weakens dramatically when data are aggregated by age group, where the correlation drops to about 0.143. This suggests that while fat is a dominant driver of thickness at any given anatomical point, age-related changes in the two layers follow partially independent trajectories, likely reflecting muscle atrophy and connective tissue remodeling that proceed on different schedules from fat redistribution.

Sexual dimorphism emerged clearly in the data, with males showing higher values than females across the landmark set, although the effect sizes were small. That combination of statistical significance and modest magnitude is a familiar pattern in facial anthropometry and carries a practical warning for forensic practitioners: knowing the sex of a skull helps narrow the expected tissue depths, but it does not remove the need for population- and age-appropriate reference values. The age analysis revealed a classic increase-peak-decline pattern, with thickness measurements rising through early adulthood, reaching a maximum in middle adulthood and then declining in older age groups, mirroring the well-documented trajectory of facial fat compartments and the progressive resorption of the underlying facial skeleton.

One of the most striking findings concerns symmetry. When the researchers compared measurements taken on the right and left sides of the same faces, they found an extremely high correlation, with a coefficient of determination of 0.992. For forensic facial approximation, this is reassuring news, because it means that reference values collected on one side of the head can be applied to the other with negligible loss of accuracy, and that genuine bilateral asymmetry in soft tissue depth is rare enough to be treated as an exception rather than a rule in reconstruction work.

The artificial intelligence component of the study is where the project pushes furthest beyond conventional practice. The team compared machine learning models against standard linear regression and found that the best-performing algorithms, particularly a Random Forest model, achieved coefficients of determination exceeding 0.90 within the dataset, outperforming conventional approaches and showing limited systematic bias. Random Forest methods work by building an ensemble of decision trees, each trained on random subsets of the data and features, and averaging their predictions; this makes them well suited to capturing the nonlinear interactions between age, sex, landmark position and body composition that govern facial tissue thickness, interactions that a single linear equation cannot represent.

The authors are careful to flag the central limitation of these results: the models demonstrated strong within-dataset performance, but independent external validation is required before the predictions can be trusted on new populations or imaging systems. A model that achieves an R-squared above 0.90 on the data it was trained and tested on may still degrade when confronted with scans from a different scanner, a different ancestry group or a different body mass distribution. This caution reflects a broader lesson from the application of machine learning in medical imaging, where promising internal metrics have repeatedly failed to translate without rigorous external testing.

Even with that caveat, the study delivers something the forensic community has long needed: a large, CT-derived, population-specific reference database for a northwestern Indian population, one of the most populous regions of the world that has been underrepresented in facial reconstruction literature. Most established tissue depth tables were built from cadaver measurements or smaller imaging samples of European and East Asian subjects, and applying them to other populations introduces systematic error into reconstructed faces. By integrating total soft tissue thickness and fat layer thickness across 73 landmarks, quantifying sex and age effects, and demonstrating near-perfect bilateral symmetry, the Chandigarh team has produced a foundation for more accurate, data-driven facial approximation. As AI-guided craniofacial reconstruction systems mature, datasets of this kind will be the raw material that determines whether a reconstructed face becomes a recognizable likeness or just an anatomical guess.

Subject of Research: AI-based CT measurement of craniofacial soft tissue and fat thickness for forensic facial reconstruction in a northwestern Indian population

Article Title: Artificial intelligence -based modeling and comparative evaluation of craniofacial soft tissue and subcutaneous fat thickness using ct imaging for forensic identification in a northwestern indian population

Article References: Battan, S. K., Sharma, M., Garg, M., Singh, P., verma, P., & Singla, N. (2026). Artificial intelligence -based modeling and comparative evaluation of craniofacial soft tissue and subcutaneous fat thickness using ct imaging for forensic identification in a northwestern indian population. International Journal of Legal Medicine. https://doi.org/10.1007/s00414-026-04000-y

Image Credits: AI Generated

DOI: 10.1007/s00414-026-04000-y

Keywords: forensic science, facial reconstruction, craniofacial soft tissue thickness, subcutaneous fat, computed tomography, artificial intelligence, machine learning, sexual dimorphism, aging, bilateral symmetry, forensic identification, population-specific database

To understand why this matters, it helps to look at how modern gait recognition actually works. Many of the best-performing systems do not analyze raw video at all. Instead, they extract a skeletal representation of the walking person — a set of joint positions, typically for the ankles, knees, hips, shoulders, elbows, wrists and head — tracked frame by frame across a walking sequence. This skeleton-based approach has a major practical advantage: because the model sees only the geometry of the body’s movement, it is naturally robust to external variations that confound appearance-based methods. A suspect wearing a heavy coat, carrying a bag, or walking under unusual lighting still produces a recognizable skeletal signature, even when their visual appearance has changed dramatically.

The standard tool for processing these skeletal sequences is the graph convolutional network, or GCN. In this framework, the joints of the body are treated as nodes in a graph, and the anatomical connections between them — shin to knee, forearm to elbow, and so on — are the edges. Information is then propagated across this graph layer by layer, allowing each joint’s representation to be shaped by the movement of its connected neighbors. The problem, as the authors of the new study point out, is that the graphs used by most existing methods are static. They are hand-designed from human anatomy and then frozen in place, the same fixed wiring pattern applied to every person, every walking style and every frame of every sequence.

That fixed wiring is a real constraint, because human walking involves coordination that goes well beyond direct anatomical links. When a person walks, the left arm swings in counterpoint to the right leg; the rotation of the pelvis is coupled to the motion of the shoulders; the timing of a heel strike ripples upward through the entire kinetic chain. These long-range, fine-grained dependencies between joints that are not physically connected — and that vary from person to person — carry some of the most discriminative information about an individual’s gait. A static graph built only from the skeletal diagram simply cannot represent them. The result is that conventional GCN-based systems capture the coarse choreography of walking but miss the subtle personal quirks that distinguish one walker from another.

AGSL-Gait tackles this limitation head-on by letting the network learn its own graph structure from data. The centerpiece of the framework is an adaptive adjacency matrix — the mathematical object that defines which joints influence which — that is not fixed by anatomy but learned during training. Rather than replacing the anatomical prior entirely, the method treats it as a starting point and learns a data-driven residual refinement on top of it, adjusting the connection strengths to capture the implicit, fine-grained inter-joint dependencies that the fixed skeleton misses. In effect, the network discovers for itself which pairs of joints carry the most identifying information about how a particular body moves, and strengthens those channels accordingly.

The second innovation concerns how the graph is partitioned. In graph convolution over a skeleton, a key design decision is which joints are grouped together and treated as sharing a transformation — for example, whether all joints are processed uniformly, whether they are split by body region such as torso, arms and legs, or whether finer subdivisions are used. Most prior work settles on a single partitioning strategy and sticks with it. The authors of the new study instead explore multiple partition strategies in combination, arguing that different ways of slicing up the body expose different aspects of its complex connectivity, and that optimizing over these patterns yields a richer model of how the body’s segments coordinate during locomotion. The combination of adaptive topology and flexible partitioning is what gives the framework its name: adaptive graph structure learning.

A third contribution is practical rather than algorithmic, but it may prove just as influential. The team constructed MoCap-Gait, a new multi-view motion capture dataset tailored specifically to forensic identification scenarios. Public gait datasets have historically been collected in relatively unconstrained settings, which is valuable for general benchmarking but does not always reflect the conditions under which forensic gait analysis is actually performed. By capturing walking data from multiple viewpoints using motion capture technology, and designing the collection around forensic use cases, the researchers have created a resource intended to test how well gait recognition methods hold up in the settings where they would matter most — courtrooms and criminal investigations, where the evidentiary value of a walking pattern may hinge on the reliability of the underlying algorithm.

The experimental case for AGSL-Gait rests on three distinct datasets. The first is CASIA-B, the most widely used benchmark in gait recognition, which contains walking sequences recorded under variations in clothing, carrying conditions and viewing angle. The second is OUMVLP-Pose, a large-scale pose-based dataset whose sheer volume tests whether a method scales beyond small laboratory collections. The third is the team’s own MoCap-Gait. Across all three, the researchers report that their approach achieves competitive performance, suggesting that the gains from adaptive graph learning generalize across data sources and capture conditions rather than being an artifact of one particular benchmark. The work was supported by the National Key Research and Development Program of China and by programs of the Ministry of Public Security, underscoring the applied, security-oriented motivation behind the research.

The broader significance of the study lies in a shift of philosophy that is playing out across machine learning more broadly. For years, the dominant practice in skeleton-based action and gait analysis was to encode human knowledge — anatomy, biomechanics, body-part hierarchies — directly into the architecture of the network. That approach is interpretable and reliable, but it caps performance at the limit of what human designers can anticipate. The AGSL-Gait framework represents the counter-trend: keep the human prior as a scaffold, but let the data decide the fine structure. As the authors put it in their abstract, adaptively learning graph topologies and partition patterns is crucial for uncovering discriminative features, precisely because human walking involves complex coordination that no fixed diagram can fully capture.

For the field of biometric identification, the implications are considerable. Gait is one of the few biometrics that can be acquired at a distance, without cooperation from the subject, and it is difficult to disguise — a person can change their face with a mask or their fingerprints with a covering, but altering the way they walk without breaking their stride is far harder. Systems that can extract more of the identifying information latent in skeletal motion, while remaining robust to clothing, carried objects and viewpoint changes, could strengthen surveillance, access control and forensic investigation alike. At the same time, the forensic orientation of the work is a reminder that accuracy claims will eventually be scrutinized in the most demanding forum of all: as evidence. With MoCap-Gait now available and adaptive graph learning demonstrating its value across three benchmarks, the study offers both a technical advance and a testing ground for the next generation of gait recognition systems.

Subject of Research: Adaptive graph structure learning with graph convolutional networks for skeleton-based gait recognition and forensic identification

Article Title: Adaptive graph structure learning for skeleton-based gait recognition

Article References: Liu, Z., Guan, L., Feng, L., Zhang, S., Bao, H., Jiang, X., Zhao, X., & Zheng, N. (2026). Adaptive graph structure learning for skeleton-based gait recognition. Complex & Intelligent Systems. https://doi.org/10.1007/s40747-026-02524-9

Image Credits: AI Generated

DOI: 10.1007/s40747-026-02524-9

Keywords: gait recognition, graph convolutional networks, adaptive adjacency matrix, skeleton-based biometrics, forensic identification, motion capture, CASIA-B, OUMVLP-Pose, MoCap-Gait dataset, graph partitioning, machine learning, biometric security