Ensemble UI: vote summary + p(ASD) histogram + collapsed per-site details

app.py

@@ -546,21 +546,80 @@ def run_gcn(file_path):
   <div><div style="font-size:0.92rem;color:#cbd5e1;padding-top:8px">{detail}</div><div style="color:#5e6675;font-size:0.7rem;margin-top:2px">Ensemble vote</div></div>
 </div></div>"""
 
-    # ── Ensemble ──
-    rows = ""
+    # ── Ensemble (vote summary + histogram + collapsed details) ──
+    asd_votes = sum(1 for _, p in per_model if p > 0.5)
+    tc_votes  = n_models - asd_votes
+    asd_pct   = (asd_votes / n_models) * 100 if n_models else 0
+    tc_pct    = 100 - asd_pct
+
+    p_values = sorted(p for _, p in per_model)
+    p_min, p_max = (p_values[0], p_values[-1]) if p_values else (0.0, 0.0)
+
+    majority_asd   = asd_votes >= tc_votes
+    maj_count      = asd_votes if majority_asd else tc_votes
+    maj_clr        = "#ef4444" if majority_asd else "#22c55e"
+    maj_lbl        = "ASD" if majority_asd else "TC"
+
+    # Histogram: bin p values into 10 buckets of width 0.1
+    bins = [0] * 10
+    for p in p_values:
+        bins[min(int(p * 10), 9)] += 1
+    max_bin = max(bins) or 1
+    hist_bars = ""
+    for i, c in enumerate(bins):
+        h    = max(int(c / max_bin * 36), 2 if c > 0 else 1)
+        clr  = "#ef4444" if i >= 5 else "#22c55e"
+        op   = 0.9 if c > 0 else 0.12
+        hist_bars += (
+            f'<div style="flex:1;display:flex;flex-direction:column;justify-content:flex-end;align-items:center;height:40px">'
+            f'<div style="color:#5e6675;font-size:0.62rem;margin-bottom:2px;height:9px">{c if c>0 else ""}</div>'
+            f'<div style="background:{clr};opacity:{op};width:78%;height:{h}px;border-radius:1px"></div>'
+            f'</div>'
+        )
+
+    # Per-site detail rows (collapsed)
+    detail_rows = ""
     for site, p in per_model:
         lbl = "ASD" if p > 0.5 else "TC"
         clr = "#ef4444" if p > 0.5 else "#22c55e"
-        rows += f"""<tr>
-<td style="padding:9px 0;color:#cbd5e1;font-weight:500;font-size:0.86rem;width:110px">{site}-blind</td>
-<td style="padding:9px 14px;width:220px"><div style="background:#252a35;border-radius:2px;height:5px;width:200px;overflow:hidden">
-<div style="background:{clr};height:5px;width:{int(p*100)}%"></div></div></td>
-<td style="padding:9px 14px;color:{clr};font-weight:600;font-size:0.85rem;width:50px">{lbl}</td>
-<td style="padding:9px 0;color:#8b95a7;font-size:0.84rem;font-variant-numeric:tabular-nums">p = {p:.3f}</td></tr>"""
+        detail_rows += f"""<tr>
+<td style="padding:6px 0;color:#cbd5e1;font-size:0.82rem;width:110px">{site}-blind</td>
+<td style="padding:6px 14px;width:200px"><div style="background:#252a35;border-radius:2px;height:4px;width:180px;overflow:hidden">
+<div style="background:{clr};height:4px;width:{int(p*100)}%"></div></div></td>
+<td style="padding:6px 14px;color:{clr};font-weight:600;font-size:0.78rem;width:40px">{lbl}</td>
+<td style="padding:6px 0;color:#8b95a7;font-size:0.78rem;font-variant-numeric:tabular-nums">p = {p:.3f}</td></tr>"""
 
     ensemble = f"""<div style="background:#161922;border:1px solid #252a35;border-radius:8px;padding:18px 24px;margin-top:10px">
-<div style="font-size:0.65rem;color:#8b95a7;letter-spacing:2px;text-transform:uppercase;margin-bottom:12px;font-weight:500">Leave-One-Site-Out Ensemble</div>
-<table style="width:100%;border-collapse:collapse">{rows}</table>
+<div style="font-size:0.65rem;color:#8b95a7;letter-spacing:2px;text-transform:uppercase;margin-bottom:14px;font-weight:500">Cross-Site Consensus · {n_models} Site-Blind Models</div>
+
+<div style="display:flex;align-items:baseline;gap:12px;margin-bottom:10px;flex-wrap:wrap">
+  <div style="font-size:1.7rem;font-weight:700;color:{maj_clr};line-height:1;font-variant-numeric:tabular-nums">{maj_count}<span style="font-size:1rem;color:#5e6675;font-weight:500"> / {n_models}</span></div>
+  <div style="color:#cbd5e1;font-size:0.86rem">site-blind models predict <span style="color:{maj_clr};font-weight:600">{maj_lbl}</span></div>
+  <div style="margin-left:auto;color:#8b95a7;font-size:0.78rem">mean p = <span style="color:#cbd5e1;font-weight:600">{p_mean:.3f}</span> · range [{p_min:.2f}, {p_max:.2f}]</div>
+</div>
+
+<div style="background:#252a35;border-radius:3px;height:9px;width:100%;overflow:hidden;display:flex;margin-bottom:6px">
+  <div style="background:#ef4444;height:9px;width:{asd_pct}%" title="{asd_votes} predict ASD"></div>
+  <div style="background:#22c55e;height:9px;width:{tc_pct}%" title="{tc_votes} predict TC"></div>
+</div>
+<div style="display:flex;justify-content:space-between;color:#5e6675;font-size:0.7rem;margin-bottom:18px">
+  <span><span style="color:#ef4444;font-weight:600">{asd_votes}</span> ASD</span>
+  <span><span style="color:#22c55e;font-weight:600">{tc_votes}</span> TC</span>
+</div>
+
+<div style="background:#0e1015;border:1px solid #1e2330;border-radius:6px;padding:10px 14px 8px;margin-bottom:12px">
+  <div style="display:flex;align-items:flex-end;gap:2px">{hist_bars}</div>
+  <div style="display:flex;justify-content:space-between;color:#5e6675;font-size:0.62rem;margin-top:3px;padding:0 1px">
+    <span>0.0</span><span>0.25</span><span style="border-left:1px dashed #3a4150;height:6px"></span><span>0.75</span><span>1.0</span>
+  </div>
+  <div style="text-align:center;color:#5e6675;font-size:0.68rem;margin-top:4px">Distribution of p(ASD) across all {n_models} site-blind models · tight clustering = strong cross-site agreement</div>
+</div>
+
+<details style="margin-top:6px">
+  <summary style="color:#8b95a7;font-size:0.76rem;cursor:pointer;font-weight:500;padding:6px 0;user-select:none">▸ Per-site breakdown · all {n_models} models</summary>
+  <table style="width:100%;border-collapse:collapse;margin-top:6px">{detail_rows}</table>
+</details>
+
 <div style="margin-top:12px;padding-top:10px;border-top:1px solid #252a35;color:#5e6675;font-size:0.76rem">
 LOSO AUC = 0.7260 · 1,102 held-out subjects · 20 acquisition sites
 </div></div>"""