Week of 2026-05-25

Individual MRF-PSD correlations

Individual-level correlations

Shifted from pooled group-level scatter plots to a per-patient zone-stratified correlation framework, following Balu’s feedback that pooling across patients diffuses findings due to inter-patient variability.

Computed Spearman ρ per patient × band × MRF column × condition, stratified by zone (EZ / NIZ / ALL contacts).

viewof patient = Inputs.select([2, 5, 6, 7, 9, 10, 11, 13], { label: "Patient (ith#)" })
viewof subset = Inputs.radio(["t1_raw", "t1_vep", "t2_raw", "t2_vep"], {
  label: "MRF type",
  format: x => ({ t1_raw: "T1 Raw", t1_vep: "T1 VEP z-score", t2_raw: "T2 Raw", t2_vep: "T2 VEP z-score" })[x],
  value: "t1_raw"
})

fig_date = "2026-05-26"
img_path = `../media/heatmap_${subset}_ith${patient}_${fig_date}.png`
html`<div style="text-align:center; margin-top: 1rem;">
  <img src="${img_path}"
       style="max-width:100%; border: 1px solid #e0e0e0; border-radius: 4px;"
       onerror="this.style.display='none'; document.getElementById('fig-missing-${patient}-${subset}').style.display='block'"/>
  <div id="fig-missing-${patient}-${subset}" style="display:none; color:#888; margin-top:1rem;">
    Figure not found: ${img_path}
  </div>
  <div style="margin-top:0.5rem; font-size:0.8rem; color:#999;">${img_path}</div>
</div>`

Because of the small number of electrode contacts for the EZ contacts, we have artificially inflated correlations.

Example scatterplots

As an example, I generated the actual scatterplots for patient ith11 below:

viewof mrf_type = Inputs.radio(["T1", "T2"], { label: "MRF type", value: "T1" })
viewof norm = Inputs.radio(["raw", "vep"], {
  label: "Normalization",
  format: x => ({ raw: "Raw", vep: "VEP z-score" })[x],
  value: "raw"
})
viewof tissue = Inputs.radio(["whole", "gm", "wm"], {
  label: "Tissue",
  format: x => ({ whole: "Whole contact", gm: "GM only", wm: "WM only" })[x],
  value: "whole"
})
viewof band = Inputs.select(
  ["delta", "theta", "alpha", "beta", "gamma", "high_gamma"],
  {
    label: "Frequency band",
    format: x => ({
      delta:      "δ (0.5–4 Hz)",
      theta:      "θ (4–8 Hz)",
      alpha:      "α (8–13 Hz)",
      beta:       "β (13–30 Hz)",
      gamma:      "γ (30–80 Hz)",
      high_gamma: "hγ (80–150 Hz)"
    })[x],
    value: "beta"
  }
)

scatter_date = "2026-05-26"
scatter_path = `../media/scatter_ith11_${mrf_type.toLowerCase()}_${norm}_${tissue}_${band}_${scatter_date}.png`
html`<div style="text-align:center; margin-top: 1rem;">
  <img src="${scatter_path}"
       style="max-width:480px; width:100%; border: 1px solid #e0e0e0; border-radius: 4px;"
       onerror="this.style.display='none'; document.getElementById('scatter-missing').style.display='block'"/>
  <div id="scatter-missing" style="display:none; color:#888; margin-top:1rem;">
    Figure not found: ${scatter_path}
  </div>
  <div style="margin-top:0.5rem; font-size:0.8rem; color:#999;">${scatter_path}</div>
</div>`

Again, I think the main takeaway for me is that the scatterplots look like a “cloud” without a clear pattern…