From Sacrum to Sole: Can One Sensor Estimate Metabolic Cost During Walking? | IMU Gait Analysis | Find Your Stride | Edinburgh Podiatrist

Introduction

Wearable technology continues its relentless march into sports medicine and clinical biomechanics. The latest study by Jung and colleagues proposes something remarkably ambitious: estimating walking energy expenditure using a single sacrum-mounted inertial measurement unit (IMU) combined with a biomechanically-informed machine learning model.

For podiatrists, running coaches, rehabilitation specialists, and sports clinicians, the promise is attractive. If metabolic cost can be estimated accurately from one wearable sensor, we move closer to real-world monitoring of gait efficiency, rehabilitation progress, and athletic performance without expensive laboratory equipment.

But does this study deliver on that promise? The answer is both yes and no.

Can one sensor estimate metabolic cost during gait analysis?

What Did the Researchers Do?

The investigators developed a machine learning framework that uses data from a single IMU positioned over the sacrum to estimate lower-limb joint mechanics and, ultimately, walking energy expenditure (EE) during gait analysis. Rather than directly predicting energy expenditure from accelerometer signals, the authors first estimated ankle, knee, and hip joint power during gait.

They then applied established biomechanical principles regarding muscle efficiency and mechanical work to derive energy expenditure estimates. This is an important distinction.

Many wearable systems rely on “black box” algorithms. The current approach is more physiologically grounded, attempting to model the mechanisms underlying energy cost rather than simply identifying patterns in data. The study included:

• 8 participants for biomechanical modelling

• 9 healthy male adults for model training

• 13 healthy male adults for independent testing

• Walking speeds ranging from 1.0 to 1.75 m/s

• Indirect calorimetry as the reference standard for energy expenditure measurement

The Most Interesting Finding for Podiatrists

Perhaps the most clinically relevant discovery was that the stance leg contributed approximately 85% of total positive whole-body mechanical power during walking.

This finding reinforces what podiatrists and gait specialists already appreciate:

Walking efficiency is fundamentally driven by what happens during stance.

The authors demonstrated exceptionally strong correlations between stance-leg sagittal plane power and whole-body energy expenditure (R = 0.92–0.98 depending on the model used). For clinicians, this supports the long-standing emphasis on:

• Propulsion mechanics

• Ankle plantarflexor function

• First metatarsophalangeal joint efficiency

• Achilles tendon energy storage and return

• Foot and ankle contribution to gait economy

• 
  

In practical terms, interventions that improve stance-phase mechanics may have disproportionate effects on overall metabolic efficiency.

Why This Matters for Running Performance

Although the study examined walking rather than running, several concepts transfer directly to endurance sport. Running economy remains one of the strongest predictors of endurance performance. Athletes with lower energy cost at a given speed generally outperform less economical runners. Traditionally, assessing running economy requires:

• Metabolic carts

• Laboratory testing

• VO₂ measurements

• Highly controlled conditions

If future versions of this technology can accurately estimate running energy expenditure in the field, it could potentially allow:

• Monitoring of running economy during training

• Assessment of footwear interventions

• Evaluation of orthotic effectiveness

• Tracking of rehabilitation progress after injury

• Detection of inefficient movement patterns before injury develops

  
  

That would represent a substantial advance in sports performance monitoring.

However, there is a significant caveat.

The Biggest Limitation: This Was Not a Running Study

Many readers will inevitably ask: “Can this be used for runners?” At present, the answer is no. The authors tested:

• Healthy adults

• Treadmill walking

• Constant speeds

• Level surfaces

  
  

They did not investigate:

• Running

• Sprinting

• Trail running

• Hill running

• Fatigued athletes

• Competitive runners

• Injured populations

  
  

The mechanics of running differ substantially from walking. Running relies more heavily on:

• Elastic tendon storage

• Stretch-shortening cycle efficiency

• Flight phases

• Different joint power distributions

• Greater contribution from the Achilles tendon and foot-ankle complex

  
  

As a result, the excellent correlations reported here cannot simply be extrapolated to running performance. The authors acknowledge this limitation themselves and call for future validation across more complex movement conditions.

A Methodological Strength: Biomechanics Before Machine Learning

One aspect of this paper deserves particular praise. Many modern AI studies throw large datasets into complex neural networks and hope useful outputs emerge. Jung et al. took the opposite approach. They first established biomechanical relationships between:

• Joint power

• Muscle efficiency

• Mechanical work

• Energy expenditure

  
  

Only then did they apply machine learning to estimate those biomechanical variables.

This makes the model more interpretable and potentially more clinically meaningful.

For sports medicine professionals increasingly confronted with AI-based technologies, this represents a more trustworthy pathway than purely data-driven prediction models.

The Sample Size Problem

Despite the innovation, clinicians should be cautious.

The independent validation cohort consisted of only 13 healthy males.

This raises several questions:

• Would the model work equally well in females?

• What about older adults?

• How would obesity influence performance?

• Would foot pathology alter accuracy?

• What happens in runners with asymmetrical gait patterns?

  
  

Small studies frequently report impressive accuracy that diminishes when tested in larger, more diverse populations. The remarkably low variability reported in this paper may partly reflect the homogeneity of the participants.

Where Are the Feet?

From a podiatric perspective, another limitation stands out. The model focuses heavily on proximal kinematics measured at the sacrum. While ankle, knee, and hip power are estimated, the system never directly measures:

• Foot motion

• Rearfoot mechanics

• Midfoot function

• Forefoot loading

• Plantar pressure distribution

  
  

For many clinical questions, especially those involving injury risk or orthotic prescription, foot-specific information remains essential. A sacral IMU may estimate overall energy expenditure effectively, but it cannot replace detailed foot and ankle assessment.

What Could This Mean for Sports Injury Management?

The future applications are intriguing. If validated in athletic populations, similar systems could potentially help clinicians monitor:

Achilles Tendinopathy

Changes in stance-phase power generation may reflect altered tendon loading patterns.

Plantar Fasciopathy

Reduced propulsion efficiency may become detectable before symptoms worsen.

Stress Fracture Rehabilitation

Energy expenditure metrics could help guide return-to-running progression.

Post-Surgical Recovery

Objective monitoring of gait efficiency could supplement traditional outcome measures.

These possibilities remain speculative, but the underlying concept is promising.

Clinical Takeaway

This study represents an important step toward practical, wearable energy expenditure monitoring. The strongest message for podiatrists and sports clinicians is not that one sensor can estimate calories burned. Rather, it is that stance-phase lower-limb mechanics appear to be an extraordinarily powerful indicator of whole-body energy cost. That finding aligns closely with contemporary understanding of:

• Running economy

• Walking efficiency

• Lower-limb injury rehabilitation

• Foot and ankle function

  
  

The technology itself is not yet ready to transform sports performance assessment or podiatric practice. However, the biomechanical principles underlying the work are sound, and the study provides a compelling roadmap for future research. For now, clinicians should view this paper as an exciting proof-of-concept rather than a practice-changing breakthrough.

Find Your Stride!

Citation

Jung J, Lim H, Jeong H, Upadhye S, Kim JH, Park S. Estimation of walking energy expenditure using a single sacrum-mounted IMU based on biomechanically-informed machine learning. Scientific Reports. 2025;15:44933. doi:10.1038/s41598-025-28393-9.