Beyond gait speed: a multidimensional motor signature of Motoric Cognitive Risk syndrome identified through domain-specific anomaly detection

This study identifies a multidimensional motor signature of Motoric Cognitive Risk (MCR) syndrome by using domain-specific anomaly detection to reveal that, unlike healthy older adults with similarly slow gait, individuals with MCR exhibit distinct deviations in gait fluctuation amplitude and temporal structure, offering potential biomarkers for early detection of neurocognitive disorders.

The Gist

Imagine your body's movement system as a highly sophisticated orchestra. For years, doctors have used a very simple way to check if this orchestra is in trouble: they just listen to the tempo (how fast the music is played). If the tempo is too slow, they worry the musicians might be losing their coordination or that the conductor (the brain) is having trouble. This "slow tempo" check is what we call gait speed, and it's the main clue doctors use to spot a condition called Motoric Cognitive Risk (MCR).

However, the problem with only listening to the tempo is that a slow song could be played by a tired musician, a broken instrument, or a confused conductor. You can't tell why it's slow just by the speed.

The New Approach: Listening to the Whole Symphony

This study decided to stop just counting the beats per minute. Instead, they put on "super-hearing" headphones to listen to the entire orchestra in detail. They wanted to see if the musicians were playing the right notes, if the rhythm was steady, and if the music was complex and lively, or if it sounded robotic and chaotic.

Here is how they did it, broken down simply:

1. The Three Groups of Musicians
The researchers gathered 97 older adults and split them into three groups:

• The Healthy Orchestra (HOA): People who walk normally and think clearly.
• The Slow but Healthy Orchestra (sHOA): People who walk slowly but have no brain fog or cognitive issues.
• The MCR Orchestra: People who walk slowly and have subjective memory complaints (they feel like their memory is slipping).

2. The "Anomaly Detector"
Think of the "Healthy Orchestra" group as the Gold Standard Sheet Music. The researchers used a smart computer program (an anomaly detector) to learn exactly what a "perfect" walk looks like based on this group.

Then, they watched the other two groups walk on a treadmill. The computer didn't just ask, "Are they slow?" It asked, "How different is their performance from the Gold Standard Sheet Music?" It looked at ten different "sections" of the orchestra, like:

• The Pace: How fast they step.
• The Rhythm: How steady the beat is.
• The Balance: How well they stay upright.
• The Complexity: How much the movement changes and adapts (like a jazz solo) versus being stiff and repetitive (like a robot).

3. The Big Discovery
Here is the "Aha!" moment of the study:

• The Slow but Healthy Group: As expected, they were "out of tune" with the Gold Standard because they walked slowly. Their pace and steps were different, but the quality of their movement was still healthy. They were just playing the song slowly.
• The MCR Group: They were also slow, but they had extra problems. Their movements weren't just slow; they were messy and rigid.
  • Their steps were more variable (one step was long, the next short, like a drummer losing the beat).
  • Their bodies were less complex (moving in a stiff, predictable, robotic way rather than a fluid, adaptable way).
  • Their movements were more divergent (small errors got bigger and bigger, like a car drifting off the road).

The Takeaway

Think of it like a car engine.

• Normal walking is a car cruising smoothly at 60 mph.
• Slow walking (sHOA) is a car cruising smoothly at 30 mph. It's slow, but the engine is fine.
• MCR walking is a car cruising at 30 mph, but the engine is sputtering, the steering is stiff, and the tires are wobbling.

The study found that MCR isn't just about walking slowly. It's about a specific, multidimensional "signature" where the movement becomes unstable, rigid, and less adaptable.

Why does this matter?
In the past, if an older person walked slowly, a doctor might just say, "You're getting older, that's normal," or "You just need to walk faster." But this new method acts like a high-tech mechanic's diagnostic tool. It can spot the specific "engine trouble" (the brain-body connection issues) that predicts future dementia, even before the person has a major memory crash.

By looking at how a person moves, not just how fast, doctors might be able to catch these risks much earlier and help people stay independent for longer.

Technical Summary

1. Problem Statement

Motoric Cognitive Risk (MCR) syndrome is currently defined by the co-occurrence of subjective cognitive complaints and slow gait speed. While this definition identifies older adults at risk for major neurocognitive disorders (NCDs), it suffers from a critical limitation: gait speed is a composite output. It reflects the aggregate result of heterogeneous neuromusculoskeletal and cognitive processes, making it a non-specific clinical marker. Consequently, it is difficult to distinguish whether slow gait in an older adult is solely due to physical frailty or if it indicates the specific underlying pathology associated with MCR. The study aims to refine the motor signature of MCR by moving beyond a single scalar metric (speed) to identify domain-specific gait deviations that uniquely characterize the syndrome.

2. Methodology

The study employed a domain-specific anomaly detection framework to quantify gait deviations relative to a normative baseline.

• Participants: The cohort consisted of 97 adults aged $\ge$ 55 years, divided into three distinct groups:
  1. MCR Group ($n=20$): Older adults with subjective cognitive complaints and slow gait.
  2. Slow Healthy Older Adults (sHOA; $n=20$): Matched to the MCR group for age and gait speed but without cognitive complaints.
  3. Healthy Older Adults (HOA; $n=57$): The normative reference group with normal gait speed and no cognitive complaints.
• Data Collection: Participants performed two 3-minute treadmill walking bouts at their preferred speed.
• Feature Engineering: Gait data was processed into ten functional gait domains, categorized into three conceptual groups:
  • Gait Pattern: Pace, rhythm, phases, postural control, symmetry.
  • Fluctuation Amplitude: Variability.
  • Temporal Structure of Fluctuations: Regulation, signal complexity, divergence of movement trajectories, attractor complexity.
  • Note: Both linear spatiotemporal variables and nonlinear trunk acceleration-derived variables were utilized.
• Statistical Modeling:
  • A Gaussian Mixture Model (GMM) was trained exclusively on the HOA data to establish a normative reference space for each of the ten domains.
  • Anomaly Scores were calculated for every individual in the sHOA and MCR groups based on their deviation from this normative space.

3. Key Contributions

• Shift from Scalar to Multidimensional Analysis: The study moves beyond the traditional reliance on gait speed alone, proposing a multidimensional motor signature that captures the complexity of gait control.
• Novel Anomaly Detection Framework: It introduces a GMM-based approach to define "normal" gait in high-dimensional space, allowing for the quantification of deviations in specific functional domains rather than global performance.
• Differentiation of Pathology: The methodology successfully distinguishes between "slow gait due to physical factors" (sHOA) and "slow gait due to MCR pathology" by isolating specific domain deviations.

4. Key Results

The analysis revealed distinct patterns of deviation across the three groups:

• Common Deviations (MCR and sHOA): Both the MCR and sHOA groups showed significantly higher anomaly scores in Gait Pattern domains (specifically pace and phases) compared to the HOA group. This confirms that both groups share the characteristic of slower, altered gait mechanics.
• Unique MCR Signature: Crucially, only the MCR group exhibited significant additional deviations in domains related to Fluctuation Amplitude and Temporal Structure.
  • Increased Variability: MCR participants showed higher step-to-step variability.
  • Altered Dynamics: Their trunk acceleration fluctuations were characterized as:
    • More divergent: Trajectories of movement separated more rapidly from the norm.
    • More predictable: Suggesting a loss of adaptive flexibility (rigidity).
    • Less complex: Indicating a reduction in the fractal complexity of movement, often associated with neurological decline.

5. Significance and Implications

• Early Detection Biomarkers: The study identifies a specific multidimensional motor signature for MCR that is distinct from simple physical slowing. These domain-specific anomaly scores serve as potential individualized biomarkers.
• Clinical Specificity: By isolating deviations in temporal structure and signal complexity, clinicians can better differentiate MCR (a precursor to dementia) from general physical frailty, even when gait speeds are identical.
• Monitoring Potential: This approach offers a quantitative, interpretable tool for monitoring the progression of neurocognitive risk in older adults, potentially enabling earlier intervention before major NCDs manifest.

In conclusion, the paper demonstrates that MCR is not merely defined by slow walking but by a specific degradation in the complexity and regulation of gait dynamics, which can be rigorously quantified using anomaly detection against a normative reference.