A question on extending OpenCap for seated posture analysis: modeling ischial COP as an outcome variable of upper-body alignment

[malchus5450-max]

Hello OpenCap team,

Thank you for sharing OpenCap as a rigorous and open platform for markerless human movement analysis. While reviewing the OpenCap pipeline, I have been particularly interested in its potential to extend beyond gait and performance analysis toward posture assessment from a prevention and recovery perspective.

With this in mind, I would like to raise a methodological question and propose a conceptual approach related to seated posture and load distribution analysis.

Background and motivation

In seated postures, I consider ischial center of pressure (COP) and left–right asymmetry to be meaningful indicators of postural imbalance and long-term neuromuscular adaptation. However, I do not view ischial COP as an independent variable; rather, I understand it as an outcome variable determined by an upstream load-transfer chain.

In this context, the key posture-related metrics I consider central to the problem are as follows:

• Plantar center of pressure (COP) location and its left–right asymmetry
• Ischial (seated) COP location and its left–right asymmetry
• Head orientation, defined as the head angle relative to the Frankfurt horizontal plane
• Lumbar lordosis/kyphosis angle
• Thoracic kyphosis/lordosis angle

I do not consider these metrics as independent features. Instead, I understand them as a coupled set of variables within a load-transfer system extending from the head through the spine and pelvis to the support surfaces. In other words, the above metrics are observables for evaluation, while their values emerge as outcomes determined by higher-level alignment state variables.

More specifically, I view ischial COP as an output that primarily reflects the combined effects of the following upper-level alignment variables:

• Three-dimensional position of the head center of mass relative to the superior endplate center of L5
• Magnitude of thoracic kyphosis and lumbar lordosis
• Three-dimensional orientation of the line connecting L5 and C1
• Three-dimensional pelvic orientation
• Foot placement and ground support conditions

From this perspective, the primary reason why seated COP estimation often lacks reliability does not lie in variations in seat material or cushioning, but rather in the fact that these dominant upper-body alignment variables are not sufficiently incorporated into the model.

Technical questions

As I understand it, OpenCap already estimates full-body kinematics and segment alignment from video data. In this regard, I would like to ask the following questions:

1. From the OpenCap perspective, how feasible would it be to treat ischial COP not as a quantity to be directly predicted, but as an outcome variable derived through calibration using limited pressure-sensor data (e.g., pressure mats or force plates)?

2. Rather than estimating COP or GRF-related quantities primarily from lower-limb kinematics, have you explored or considered extension structures in which these quantities are explicitly constrained by spinal curvature (thoracic/lumbar) and head center-of-mass position?

3. In your view, what is the most significant bottleneck in generalizing seated COP estimation within a markerless framework?
   In particular, do you think that incorporating upper-body alignment variables (e.g., spinal curvature and head center of mass) into the model structure could help alleviate this bottleneck?
   (For example, limitations in observability, insufficient training data, or assumptions embedded in the model structure.)

Hypothesis

I hypothesize that, by incorporating the upper-level alignment variables related to COP described above, it would be possible to significantly reduce systematic error in left–right asymmetry of seated COP, even if limitations remain in absolute COP accuracy.

I would greatly appreciate your thoughts on whether this approach is conceptually consistent with OpenCap’s observation and modeling philosophy, and what practical limitations or considerations you foresee from an implementation standpoint.

Thank you very much for your time and for making this excellent work openly available.

Best regards,
Eam Taekyoung
Independent researcher with long-term interest in posture and human alignment