BrainMetastabilityAnalyzerTool/index.html at main · anttiluode/BrainMetastabilityAnalyzerTool · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<meta name="description" content="2026 reality audit of Φ-Dwell: spatial sensor-phase modes, raw and cleaned Alzheimer EEG results, spectral slowing comparison, and frozen external validation gate." />
<title>Φ-Dwell Reality Audit — Brain Metastability Analyzer</title>
<style>
:root{--bg:#071018;--panel:#0d1a24;--panel2:#112330;--text:#e8f0f5;--muted:#9fb0bb;--line:#243947;--good:#65d5a5;--warn:#f4c76f;--bad:#ff7f7f;--cyan:#68c7e8;--purple:#b99cf3;--max:1120px}
*{box-sizing:border-box}html{scroll-behavior:smooth}body{margin:0;background:radial-gradient(circle at 18% 0%,rgba(104,199,232,.10),transparent 32rem),radial-gradient(circle at 85% 12%,rgba(185,156,243,.08),transparent 28rem),var(--bg);color:var(--text);font-family:Inter,ui-sans-serif,system-ui,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif;line-height:1.55}a{color:var(--cyan)}.wrap{max-width:var(--max);margin:0 auto;padding:0 22px}nav{position:sticky;top:0;z-index:10;background:rgba(7,16,24,.88);backdrop-filter:blur(14px);border-bottom:1px solid var(--line)}nav .wrap{display:flex;gap:18px;align-items:center;min-height:58px}nav strong{margin-right:auto}nav a{color:var(--muted);text-decoration:none;font-size:14px}nav a:hover{color:var(--text)}header{padding:84px 0 56px}.kicker{color:var(--cyan);font-weight:800;text-transform:uppercase;letter-spacing:.14em;font-size:12px}h1{margin:12px 0 16px;font-size:clamp(42px,7vw,78px);line-height:.98;letter-spacing:-.045em;max-width:980px}h2{font-size:clamp(30px,4vw,46px);line-height:1.08;margin:0 0 14px;letter-spacing:-.03em}h3{margin:0 0 8px;font-size:20px}p{margin:0 0 16px}.lead{max-width:900px;color:#c6d4dc;font-size:clamp(18px,2.2vw,24px)}.muted{color:var(--muted)}.badge-row{display:flex;flex-wrap:wrap;gap:10px;margin-top:28px}.badge{border:1px solid var(--line);background:rgba(17,35,48,.7);padding:7px 11px;border-radius:999px;color:var(--muted);font-size:13px}.buttons{display:flex;flex-wrap:wrap;gap:10px;margin-top:22px}.button{display:inline-block;text-decoration:none;border:1px solid var(--line);border-radius:10px;padding:10px 14px;background:var(--panel2);color:var(--text);font-weight:650}.button.primary{border-color:rgba(104,199,232,.55);background:rgba(104,199,232,.12)}section{padding:58px 0;border-top:1px solid rgba(36,57,71,.7)}.grid{display:grid;gap:16px}.g3{grid-template-columns:repeat(3,1fr)}.g2{grid-template-columns:repeat(2,1fr)}@media(max-width:800px){.g3,.g2{grid-template-columns:1fr}nav a{display:none}}.card{background:linear-gradient(180deg,rgba(17,35,48,.78),rgba(13,26,36,.78));border:1px solid var(--line);border-radius:16px;padding:20px}.card.good{border-color:rgba(101,213,165,.38)}.card.warn{border-color:rgba(244,199,111,.38)}.card.bad{border-color:rgba(255,127,127,.36)}.state{font-size:12px;letter-spacing:.12em;text-transform:uppercase;font-weight:800;margin-bottom:8px}.good .state{color:var(--good)}.warn .state{color:var(--warn)}.bad .state{color:var(--bad)}code,pre{font-family:"SFMono-Regular",Consolas,"Liberation Mono",monospace}pre{white-space:pre-wrap;background:#08131b;border:1px solid var(--line);border-radius:14px;padding:18px;color:#cfe0e9;overflow-x:auto}.pipeline{font-size:clamp(15px,2vw,19px);text-align:center;color:#d8e6ed}.metric{font-size:31px;font-weight:760;letter-spacing:-.03em}.metric small{font-size:13px;color:var(--muted);font-weight:500}.callout{border-left:4px solid var(--cyan);padding:4px 0 4px 18px;margin:24px 0;font-size:20px;color:#d8e7ee;max-width:950px}.callout.warnline{border-color:var(--warn)}table{width:100%;border-collapse:collapse;font-size:14px}th,td{padding:12px 10px;text-align:left;border-bottom:1px solid var(--line);vertical-align:top}th{color:#b9c9d1;font-size:12px;letter-spacing:.08em;text-transform:uppercase}.gate{display:grid;grid-template-columns:170px 1fr;gap:16px;padding:14px 0;border-bottom:1px solid var(--line)}.gate b{color:var(--cyan)}footer{padding:44px 0 70px;color:var(--muted);font-size:13px;border-top:1px solid var(--line)}
</style>
</head>
<body>
<nav><div class="wrap"><strong>Φ-Dwell · 2026 Audit</strong><a href="#core">Core</a><a href="#result">Result</a><a href="#slowing">Slowing</a><a href="#severity">Severity</a><a href="#gate">External gate</a></div></nav>
<header class="wrap">
<div class="kicker">BrainMetastabilityAnalyzerTool</div>
<h1>The group effect survived cleaning. The diagnostic advantage did not.</h1>
<p class="lead">Φ-Dwell remains an interesting spatial sensor-phase representation of EEG dynamics. But after recomputing the frozen feature on cleaned derivative EEG, ordinary spectral slowing explains the useful AD-versus-control discrimination just as well.</p>
<div class="badge-row"><span class="badge">88/88 derivative subjects processed</span><span class="badge">dwell AD/CN p = 0.0062</span><span class="badge">spectral AUC = 0.778</span><span class="badge">spectral + dwell AUC = 0.777</span><span class="badge">increment = −0.001</span><span class="badge">not a clinical diagnostic tool</span></div>
<div class="buttons"><a class="button primary" href="INTERNAL_DERIVATIVE_AUDIT_RESULT_2026.md">Cleaned EEG receipt</a><a class="button" href="AUDIT_2026.md">Full audit</a><a class="button" href="https://github.com/anttiluode/BrainMetastabilityAnalyzerTool">Repository</a></div>
</header>
<section id="core"><div class="wrap"><div class="kicker">What survives</div><h2>Keep the measurement. Drop the metaphysics.</h2><p class="muted">The graph Laplacian is built from electrode geometry. Its eigenvectors are therefore modes of the <b>sensor layout</b>, not structural-connectome modes and not a holographic reconstruction.</p><pre class="pipeline">EEG phase at sensors
↓
project onto graph-Laplacian sensor modes
↓
dominant spatial mode in δ / θ / α / β / γ
↓
dwell time · transitions · multi-band states</pre><div class="grid g3"><div class="card good"><div class="state">Keep</div><h3>Spatial phase modes</h3><p class="muted">A deterministic transform of multichannel phase organization.</p></div><div class="card good"><div class="state">Keep</div><h3>Dwell dynamics</h3><p class="muted">How long each band remains in the same dominant sensor mode.</p></div><div class="card warn"><div class="state">Frozen candidate</div><h3>Dwell gradient</h3><p class="muted">Slope of log dwell across delta → theta → alpha → beta → gamma.</p></div></div></div></section>
<section id="result"><div class="wrap"><div class="kicker">Raw versus cleaned</div><h2>The crucial result changed after preprocessing.</h2><p>In the raw-style representation, adding dwell gradient to age + three simple spectral features increased mean internal CV AUC by about <b>+0.039</b>. After recomputing the same frozen dwell feature on derivative / cleaned EEG, that increment became essentially zero.</p><table><thead><tr><th>Quantity</th><th>Raw-style</th><th>Derivative / cleaned</th></tr></thead><tbody><tr><td>dwell AD-CN p</td><td>0.000277</td><td>0.006175</td></tr><tr><td>Model A · age + spectral</td><td>0.729 AUC</td><td>0.778 AUC</td></tr><tr><td>Model B · age + dwell</td><td>0.756 AUC</td><td>0.694 AUC</td></tr><tr><td>Model C · age + spectral + dwell</td><td>0.768 AUC</td><td>0.777 AUC</td></tr><tr><td><b>C − A</b></td><td><b>+0.039</b></td><td><b>−0.001</b></td></tr></tbody></table><div class="callout">The univariate disease association survives cleaning. The claim that Φ-Dwell adds useful Alzheimer discrimination beyond ordinary slowing does not survive this internal robustness check.</div><div class="grid g3"><div class="card warn"><div class="metric">0.0062 <small>p</small></div><h3>Dwell still differs</h3><p class="muted">AD mean −0.3551 versus CN −0.3275.</p></div><div class="card good"><div class="metric">0.778 <small>AUC</small></div><h3>Spectral baseline</h3><p class="muted">Age + alpha power + theta/alpha + peak alpha frequency.</p></div><div class="card bad"><div class="metric">−0.001 <small>AUC</small></div><h3>Dwell increment</h3><p class="muted">No improvement after cleaning.</p></div></div></div></section>
<section id="slowing"><div class="wrap"><div class="kicker">The boring explanation</div><h2>Alzheimer's spectral slowing is not background theory here. It is in the data.</h2><table><thead><tr><th>Cleaned feature</th><th>AD mean</th><th>CN mean</th><th>p</th></tr></thead><tbody><tr><td>alpha relative power</td><td>0.0494</td><td>0.0794</td><td>0.002351</td></tr><tr><td>theta / alpha ratio</td><td>2.5318</td><td>1.6045</td><td>0.0000732</td></tr><tr><td>peak alpha frequency</td><td>7.493 Hz</td><td>8.681 Hz</td><td>0.000155</td></tr></tbody></table><p class="muted" style="margin-top:18px">These simple measures already capture strong disease-related slowing. The current evidence therefore favors Φ-Dwell as an alternative spatial-dynamical view of the same disease physiology rather than a new diagnostic signal.</p><div class="callout warnline">Do not retune bands, graph modes, dwell definition or gradient direction on ds004504 to recover the lost +0.039. That would turn a clean robustness result into post-hoc rescue.</div></div></section>
<section id="severity"><div class="wrap"><div class="kicker">Severity claim</div><h2>The historical pooled MMSE story remains dead.</h2><div class="grid g2"><div class="card bad"><div class="state">Within AD</div><div class="metric">ρ 0.269</div><p class="muted">p = 0.113, n = 36 after cleaning.</p></div><div class="card bad"><div class="state">Within FTD</div><div class="metric">ρ −0.009</div><p class="muted">p = 0.969, n = 23.</p></div></div><p class="muted" style="margin-top:18px">The old pooled correlation mixed diagnosis with MMSE. The current data do not support saying that dwell gradient tracks cognitive severity.</p></div></section>
<section id="age"><div class="wrap"><div class="kicker">A useful negative</div><h2>The scary control-age correlation was also preprocessing-sensitive.</h2><p>Raw-style controls showed dwell versus age at <code>rho = −0.599, p = 0.000597</code>. In the cleaned derivative representation it became <code>rho = −0.100, p = 0.607</code>. That argues against treating the earlier age effect as a stable biological property.</p></div></section>
<section id="gate"><div class="wrap"><div class="kicker">Frozen next gate</div><h2>New people, unchanged feature.</h2><p>The only decisive Alzheimer test now is an independent cohort. The cleaned internal result makes the prediction harder and therefore more useful.</p><div class="gate"><b>Primary feature</b><div><code>dwell_gradient</code> unchanged.</div></div><div class="gate"><b>Model A</b><div>Age + alpha relative power + theta/alpha ratio + peak alpha frequency.</div></div><div class="gate"><b>Model B</b><div>Age + dwell gradient.</div></div><div class="gate"><b>Model C</b><div>Age + spectral features + dwell gradient.</div></div><div class="gate"><b>Primary comparison</b><div>C versus A on completely independent AD/CN subjects.</div></div><div class="gate"><b>No rescue</b><div>No changing bands, graph sigma, number of modes, word step, log transform, gradient direction or age handling after labels are seen.</div></div><div class="grid g3" style="margin-top:24px"><div class="card bad"><div class="state">External null</div><h3>Retire biomarker claim</h3></div><div class="card warn"><div class="state">Replicates only</div><h3>Interesting representation, no new biomarker</h3></div><div class="card good"><div class="state">Incremental signal</div><h3>Worth serious follow-up</h3></div></div></div></section>
<section id="context"><div class="wrap"><div class="kicker">Context</div><h2>Why the baseline matters</h2><p class="muted">Recent dementia EEG literature repeatedly reports spectral slowing, lower dominant/alpha frequency and altered theta/alpha balance. Any proposed low-cost EEG biomarker should therefore demonstrate added value beyond those simple measures.</p><p><a href="https://pubmed.ncbi.nlm.nih.gov/40379988/">2025 systematic review of electrophysiological dementia markers</a> · <a href="https://pubmed.ncbi.nlm.nih.gov/39449097/">spectral and connectivity features in AD</a> · <a href="https://pubmed.ncbi.nlm.nih.gov/40013094/">2025 EEG connectivity review</a></p></div></section>
<footer><div class="wrap"><p><b>Clinical boundary:</b> research software only. No prospective diagnostic accuracy, clinical utility, or generalization across acquisition systems has been established.</p><p>2026 reality audit · BrainMetastabilityAnalyzerTool · MIT</p></div></footer>
</body>
</html>