Interactions between age and sex in multiscale entropy and spectral power changes across the lifespan

This study validates that aging is associated with a shift toward fine-scale entropy and specific spectral changes in the brain, while demonstrating that these patterns diverge by sex and that entropy-based measures capture unique aspects of temporal organization not reflected by spectral power alone.

The Gist

The Big Picture: Listening to the Brain's "Chatter"

Imagine your brain is a massive, bustling city. Every time you think, feel, or remember something, it's like a conversation happening between millions of people (neurons) across different neighborhoods.

For a long time, scientists have studied this city by looking at how loud the conversations are. They measure the volume of specific types of chatter (like the "Alpha" or "Beta" waves). This is called Spectral Power. It's like measuring the decibels of a radio station.

However, this new study suggests that volume isn't the whole story. Sometimes, two radio stations can have the same volume, but one is playing a chaotic, unpredictable jazz improvisation, while the other is playing a boring, repetitive loop. The pattern of the sound matters just as much as the volume.

This study uses a tool called Multiscale Entropy (MSE). Think of MSE as a "Complexity Meter." It doesn't just measure how loud the brain is; it measures how interesting, varied, and organized the brain's signals are over different lengths of time.

The Main Characters: Age and Sex

The researchers looked at data from nearly 600 healthy people ranging from young adults (18 years old) to seniors (88 years old). They wanted to see how the brain's "complexity" changes as we get older and if men and women experience this differently.

1. The Aging Effect: From a Symphony to a Solo

The Finding: As people get older, their brain signals change in a specific way.

• Young Brains: Think of a young brain like a symphony orchestra. It has a rich, complex sound that works well over long periods (coarse scales) and short bursts (fine scales). It's integrated and flexible.
• Older Brains: As people age, the brain shifts. It becomes less like a symphony and more like a soloist playing a very fast, repetitive riff.
  • Fine-scale entropy goes UP: The brain gets "noisier" or more chaotic in the very short, split-second moments.
  • Coarse-scale entropy goes DOWN: The brain loses its ability to maintain complex patterns over longer stretches of time.

What this means: The aging brain is still active, but it's becoming more "segregated." Instead of different parts of the brain working together in a big, coordinated network, they are starting to work in smaller, isolated pockets. It's like the orchestra members stopped playing together and started tapping their own feet independently.

2. The Sex Difference: The "Crossover" Point

The Finding: Men and women don't age the same way, and the difference becomes obvious around middle age.

• Young Adults: In their 20s and 30s, men and women have very similar brain complexity patterns.
• Middle Age (The Crossover): Around age 50 (roughly the age of menopause for women), the paths diverge.
  • Women: Start to show a sharper drop in long-term complexity and a rise in short-term "noise."
  • Men: Their brains hold onto the "symphony" pattern a bit longer, or at least change differently.

The Analogy: Imagine two cars driving down a highway. For the first 30 years, they drive side-by-side at the same speed. Around the 50-year mark, one car (women) starts to take a slightly different route, perhaps due to a change in fuel (hormones like estrogen), while the other car (men) stays on the original path for a bit longer.

The Secret Sauce: Why Use MSE?

The researchers compared their "Complexity Meter" (MSE) with the old "Volume Meter" (Spectral Power).

• The Volume Meter (PSD): This told them the brain was slowing down (like a record player slowing down) and changing its rhythm. It confirmed the aging effects.
• The Complexity Meter (MSE): This told them the same story about aging, BUT it was much better at spotting the differences between men and women.

The Key Takeaway:
Imagine you are trying to describe a painting.

• Spectral Power is like saying, "This painting has a lot of blue and some red."
• MSE is like saying, "This painting has a lot of blue, but the brushstrokes are chaotic in the corners and smooth in the center."

Even though both tools are looking at the same painting, the Complexity Meter (MSE) gave a much clearer picture of the differences between the male and female brains. It captured the "texture" of the brain's activity that the volume meter missed.

The "Linear" Surprise

One of the most interesting parts of the study was a test they did called "Phase Randomization."

• The Test: They scrambled the timing of the brain signals so that any complex, non-linear patterns were destroyed, leaving only the basic "volume" and "rhythm" (linear patterns).
• The Result: Even after scrambling the complex patterns, the results looked almost exactly the same!

What this means: The changes in brain complexity as we age are mostly driven by simple, linear changes in the brain's rhythm (like a clock ticking slower), rather than some mysterious, complex "non-linear" magic. The brain isn't becoming "broken" in a weird way; it's just shifting its rhythm in a predictable, linear way.

Summary for the Everyday Reader

1. Aging changes the brain's rhythm: As we get older, our brains become a bit more chaotic in the short term and less coordinated in the long term.
2. Men and Women age differently: Around age 50, women's brains start to show these changes more distinctly than men's, likely linked to hormonal changes.
3. New tools are better: Measuring the complexity of brain signals (MSE) is a powerful new way to see these changes. It sees the "texture" of the brain's activity, not just the volume.
4. It's a simple shift: These changes aren't due to some weird, unpredictable glitch; they are a natural, linear shift in how our brain's electrical signals are organized.

The Bottom Line: This study gives us a better map of how healthy brains change over a lifetime. By understanding these normal changes, scientists can better spot when a brain is changing in an unhealthy way (like in dementia) because they know what the "normal" aging path looks like for both men and women.

Technical Summary

1. Problem Statement

Brain aging involves complex changes in neural dynamics, yet the normative evolution of brain signal complexity across the healthy lifespan remains less understood than structural or functional neurodegeneration. Previous studies have established that aging is associated with a shift in Multiscale Entropy (MSE): an increase in fine-scale entropy (reflecting local, short-range fluctuations) and a decrease in coarse-scale entropy (reflecting long-range, integrated dynamics). This shift is interpreted as a move from integrated to segregated network organization.

However, critical gaps remain:

• Sample Size: Previous findings on MSE and aging were often derived from small samples, limiting generalizability.
• Demographic Interactions: The interplay between sex and age on MSE trajectories is unclear, with conflicting reports in the literature.
• Cognitive Correlates: It is unknown how MSE patterns interact with specific cognitive domains (fluid intelligence, visual short-term memory, generalized cognitive function) across the lifespan.
• Methodological Comparison: The relationship between MSE (a time-domain complexity measure) and Power Spectral Density (PSD, a frequency-domain measure) regarding these demographic effects is not fully elucidated. Specifically, it is unclear if MSE captures unique non-linear temporal structures or if it merely reflects linear spectral properties.

2. Methodology

Dataset and Participants:

• Source: Cambridge Centre for Ageing and Neuroscience (CamCAN) dataset (Phase 2).
• Sample: Final $N=587$ healthy adults (ages 18–88), uniformly distributed across age deciles (363 females, 283 males).
• Data: Resting-state MEG (8:40 min, eyes-closed) and structural MRI.
• Preprocessing:
  • MEG data was cleaned using temporal signal space separation (tSSS), artifact removal via Independent Component Analysis (ICA), and bandpass filtering (1–90 Hz).
  • Source localization was performed using a Boundary Element Model (BEM) and a unit-noise gain minimum variance beamformer.
  • Time series were extracted for 68 Regions of Interest (ROIs) based on the Desikan-Killiany Atlas.

Analytical Measures:

1. Multiscale Entropy (MSE): Calculated using Sample Entropy (SampEn) across multiple temporal scales (coarse-graining). The study addressed edge cases where entropy was undefined by modifying the maximal entropy calculation.
2. Power Spectral Density (PSD): Calculated using the Welch method, normalized as Z-scores within ROIs.
3. Surrogate Analysis: Phase-randomized time series were generated to isolate non-linear dynamics. If MSE changes persist after phase randomization (which destroys non-linear phase dependencies but preserves the power spectrum), the effects are driven by linear autocorrelations.
4. Behavioral Measures:
   • Age and Sex.
   • Fluid Intelligence (Cattell test, age-adjusted).
   • Visual Short-Term Memory (VSTM).
   • Generalized Cognitive Function (ACE-R).

Statistical Approach:

• Partial Least Squares (PLS): Two variants were used:
  • Mean-Centered Task PLS: To identify latent variables (LVs) distinguishing groups (Age/Sex).
  • Behavioral PLS: To identify patterns of maximal covariance between brain data (MSE/PSD) and behavioral variables.
• Validation: Significance was assessed via permutation testing (10,000 iterations) and reliability via bootstrap ratios (weight/SE > 3).
• Comparison: Cosine similarity was used to compare LVs across different models (e.g., MSE vs. PSD, grouped vs. ungrouped analyses).

3. Key Contributions

• Large-Scale Validation: Confirmed the classic "aging profile" of MSE (increased fine-scale, decreased coarse-scale entropy) in a large, open-access dataset ($N=587$).
• Sex-Age Interaction: Demonstrated that sex differences in brain signal complexity are not static but interact with age, becoming more pronounced in middle and old age.
• Decoupling of Metrics: Showed that while MSE and PSD share variance related to linear autocorrelations, MSE provides a distinct, higher-order summary of temporal organization that separates age and sex effects more cleanly than spectral power.
• Linearity of Aging Effects: Through phase-randomization, the study determined that the observed age-related MSE changes are primarily driven by linear temporal correlations (spectral structure) rather than non-linear dynamics, yet MSE still offers a unique compression of this information.

4. Key Results

A. Age-Related Changes (Replication):

• MSE: Aging was associated with increased fine-scale entropy and decreased coarse-scale entropy across the cortex.
• PSD: Aging showed classic signatures: slowing of peak alpha rhythms, increased beta-band activity, and reduced gamma-band activity.

B. Sex and Age Interactions:

• MSE: Sex differences were minimal in young adults but diverged significantly in middle-aged and older groups.
  • Older Males: Showed lower fine-scale entropy and increased mid-scale entropy compared to females.
  • Older Females: Showed higher fine-scale entropy.
  • This suggests a "crossover" point around age 50 (coinciding with menopause) where female brain signal complexity trajectories diverge from males.
• PSD: Sex differences in spectral power also increased with age but manifested differently (e.g., females showed decreased alpha/high gamma and increased beta compared to males).

C. Cognitive Correlates:

• Fluid Intelligence: Strongly correlated with the age-related MSE pattern (LV1).
• VSTM & Generalized Cognition: Correlated with the sex-related MSE pattern (LV2).
• Moderation: Contrary to some prior hypotheses, sex did not moderate the relationship between fluid intelligence and MSE; age remained the primary driver.

D. MSE vs. PSD and Non-Linearity:

• Phase Randomization: When non-linear dynamics were removed via phase randomization, the behavioral PLS patterns (LVs) remained nearly identical to the original data. This indicates that the MSE changes observed are driven by linear autocorrelations (spectral power) rather than complex non-linear interactions.
• Divergent Summaries: Despite sharing the same underlying linear variance, MSE and PSD mapped these effects differently:
  • MSE: Separated age and sex effects into distinct Latent Variables (LV1 = Age, LV2 = Sex).
  • PSD: Mixed age and sex effects across both significant LVs.
  • Conclusion: MSE acts as a "compressor" that reorganizes linear spectral information into a time-domain representation that highlights demographic distinctions (specifically sex) more effectively than raw spectral power.

5. Significance

• Biomarker Potential: MSE is validated as a sensitive, population-level biomarker for functional brain aging that captures subtle demographic nuances (specifically sex differences) that spectral power alone may obscure.
• Mechanistic Insight: The finding that MSE aging effects are linearly driven suggests that the "loss of complexity" in aging is fundamentally a change in the spectral slope and autocorrelation of neural signals, rather than a breakdown of non-linear chaotic dynamics.
• Clinical Utility: The ability of MSE to differentiate between age and sex effects, and to link specific cognitive domains to distinct neural complexity patterns, supports its use in distinguishing adaptive aging from pathological decline (e.g., dementia).
• Methodological Advancement: The study clarifies the relationship between time-domain complexity and frequency-domain power, demonstrating that while they are mathematically linked, MSE offers a unique, behaviorally relevant perspective on neural organization that complements traditional spectral analysis.