Computational Biomechanics

FRCS Level

The Principle

A joint cannot be modelled in isolation, because the forces crossing it come from muscles attached elsewhere.

The body has more muscles than it has degrees of freedom, so any required movement can be produced by many different muscle combinations.

That is the muscle redundancy problem, and it means muscle forces cannot be calculated. They have to be estimated by assuming what the body is trying to do.

At the Knee

A musculoskeletal model represents bones as segments, joints as constraints and muscles as lines of action with force capacity.

Inverse dynamics works backwards from measured motion and ground reaction force to the net moment at each joint. 3.4.3 covers those forces.

Static optimisation then distributes that net moment among muscles by minimising something, usually the sum of squared activations.

The assumption that the body minimises muscle effort is a modelling convenience with partial physiological support.

The output is where every knee contact force figure in gait literature comes from. 3.14.1 covers gait.

What Changes It

The generic model scaled to the subject, since most models start from one cadaver's anatomy.

Muscle moment arms, which drive the force estimates and vary between models.

The optimisation criterion, which is an assumption about the nervous system.

Whether co-contraction is permitted, since minimising effort tends to eliminate it. 3.14.11 covers co-contraction in arthritis.

The motion data quality feeding the inverse dynamics. 3.16.7 covers motion capture.

Clinical Relevance

Read a knee contact force figure as the output of an optimisation rather than a measurement.

Expect models to underestimate co-contraction and therefore joint compression.

Compare contact force figures within a modelling framework rather than between them.

State the model and the optimisation criterion with any contact force number.

Fellowship Level

At the Knee

Electromyography driven models replace the optimisation assumption with measured muscle activity, which changes the answer where co-contraction matters.

In the arthritic knee that difference is large, because co-contraction is raised and optimisation models do not predict it. 3.14.11 covers the gait.

Instrumented implant data has been used to test these models, and the comparison is the field's best validation. 3.14.12 covers that source.

Those comparisons found that no single model predicted contact force accurately across all subjects and activities, which was reported openly by the field.

Generic models scaled by height and mass carry the original specimen's muscle geometry, so subject specificity is limited to segment lengths.

What Changes It

Whether the model is optimisation driven or electromyography driven.

Whether subject specific geometry was used or a scaled generic model.

The activity modelled, since accuracy differs between walking and more demanding tasks.

Soft tissue artefact in the input kinematics, which propagates through everything. 3.16.7 covers it.

Joint model complexity, since a hinge knee and a six degree of freedom knee give different moment arms.

How It Is Measured

Comparison against instrumented implant measurements, which is the only direct in vivo check available.

Grand challenge competitions in which groups predicted the same subject's contact forces blind, which is unusually good practice.

Agreement between models on the same data, which is moderate and rarely reported.

Sensitivity analysis on moment arms and optimisation criteria.

The best available validation comes from replaced knees, so the native knee's contact forces remain unvalidated.

Clinical Relevance

Prefer electromyography driven estimates where co-contraction is part of the question.

Treat contact force figures for the native knee as unvalidated, since the validation data comes from implants.

Note whether the model was subject specific in geometry or only in scaling.

Expect models to agree moderately with each other and to be reported as if they agree well.

Consultant Perspective

At the Knee

This page sits behind a large share of the quantitative claims in Section 3, so the decision it forces is how much weight to take off a chapter.

On one side, the estimates are physically constrained, internally consistent and the only route to a number for a loaded living knee. 3.4.3 covers what they produce.

On the other, the one assumption that most affects them, what the nervous system is minimising, is a convenience, and the field's own blind competitions showed the consequences.

The weight moves with how the number is used. For ranking activities by knee load the estimates are serviceable. For a threshold a patient should stay below they are not.

The field's openness about its own limits is a point in its favour and is rarely carried across when its figures are quoted elsewhere.

What Changes It

Whether the use is ranking or thresholding, which is the dividing line.

Whether the question concerns a native or a replaced knee.

Whether co-contraction is central, in which case optimisation models are the wrong tool.

Whether a figure is being quoted as a number or as a direction of change.

The purpose, since a research hypothesis tolerates uncertainty a patient conversation does not.

How It Is Measured

Blind prediction against instrumented implant data, which is the field's reference and covers replaced knees.

Between model agreement, which is moderate.

Reproducibility, which is improving because several major models are open source.

Sensitivity analyses, which quantify the dependence on the assumption.

There is no validation of native knee contact force, and there is no obvious route to one that does not involve instrumenting a healthy joint.

Clinical Relevance

Use these estimates for ranking and direction and not for patient specific thresholds.

State that native knee contact forces are unvalidated when quoting one.

Prefer electromyography driven work for arthritic and co-contracting knees.

Carry the field's own stated uncertainty across when quoting its figures, since it is usually dropped.

Weigh the patient's age and physiology, comorbidities, knee demand, psychological state and their own ideas, concerns and expectations before converting a modelled load into advice, since advice derived from an estimate is still advice.