Update figure generation + regenerate paper figures

scripts/make_figures.py

@@ -365,10 +365,10 @@ def fig5_difficulty():
     for thr, lab in ((0.5, ""), (1.0, ""), (2.0, "")):
         axA.axvline(thr, color="#c8c8c8", ls=":", lw=0.9, zorder=0)
     axA.set_yticks(y)
-    axA.set_yticklabels([a for a, _ in items], fontsize=5.6)
+    axA.set_yticklabels([a for a, _ in items], fontsize=8)
     axA.set_ylim(-0.7, len(items) - 0.3)
     axA.set_xlim(0, 2.6)
-    axA.set_xlabel("Trained-rater localization error, AFLE (mm)", fontsize=8.5)
+    axA.set_xlabel("Trained-rater localization error, AFLE (mm)", fontsize=11)
     axA.plot([], [], "o", color="#555", ms=5.5, label="median")
     axA.plot(
         [], [], "x", color="#555", ms=4.5, mew=1.1, label="mean (heavy tail)"
@@ -377,9 +377,9 @@ def fig5_difficulty():
         [], [], "-", color="#999", lw=2.6, alpha=0.6, label="IQR (p25–p75)"
     )
     axA.legend(
-        fontsize=6.3, loc="lower right", frameon=False, handletextpad=0.4
+        fontsize=9, loc="lower right", frameon=False, handletextpad=0.4
     )
-    panel_letter(axA, "A", "A 3.9× difficulty spectrum across 32 landmarks")
+    panel_letter(axA, "A", "")
 
     # Panel B — the same 1.2 mm placement, two verdicts
     axB = fig.add_subplot(gs[0, 1])
@@ -417,40 +417,13 @@ def fig5_difficulty():
             markeredgecolor="white",
             markeredgewidth=0.8,
         )
-        xs, ys = _pctile_curve(s)
-        pct = float(np.interp(probe, xs, ys))  # error rank (drives the verdict)
-        p = round(100 - pct)                    # precision pct shown (higher = better)
-        suf = ("th" if 11 <= p % 100 <= 13
-               else {1: "st", 2: "nd", 3: "rd"}.get(p % 10, "th"))
-        verdict = (
-            f"within range\n(≈{p}{suf} pct, near median)"
-            if pct < 75
-            else f"outside expert range\n(≈{p}{suf} pct)"
-        )
-        axB.annotate(
-            f"{a}: {verdict}",
-            (probe, yi + 0.22),
-            fontsize=7.2,
-            ha="center",
-            va="bottom",
-            color=c,
-            fontweight="bold",
-        )
     axB.axvline(probe, color="#111111", ls="--", lw=1.3, zorder=5)
-    axB.annotate(
-        "one placement:\n1.2 mm from target",
-        (probe, -0.62),
-        fontsize=7.4,
-        ha="center",
-        va="top",
-        color="#111111",
-    )
     axB.set_yticks([0, 1])
-    axB.set_yticklabels(["LIGO\n(hard)", "AC\n(easy)"], fontsize=7.5)
+    axB.set_yticklabels(["LIGO\n(hard)", "AC\n(easy)"], fontsize=10.5)
     axB.set_ylim(-1.05, 1.7)
     axB.set_xlim(0, 4.0)
-    axB.set_xlabel("Trained-rater AFLE distribution (mm)", fontsize=8.5)
-    panel_letter(axB, "B", "The same 1.2 mm — two verdicts")
+    axB.set_xlabel("Trained-rater AFLE distribution (mm)", fontsize=11)
+    panel_letter(axB, "B", "")
 
     # Panel C — the mm→percentile calibration the platform applies
     axC = fig.add_subplot(gs[0, 2])
@@ -480,17 +453,16 @@ def fig5_difficulty():
     axC.axhline(50, color="#dddddd", lw=0.8, zorder=0)
     axC.set_xlim(0, 4)
     axC.set_ylim(0, 100)
-    axC.set_xlabel("Placement error (mm)", fontsize=8.5)
-    axC.set_ylabel("Precision percentile (vs trained raters)", fontsize=8.5)
+    axC.set_xlabel("Placement error (mm)", fontsize=11)
+    axC.set_ylabel("Precision percentile (vs trained raters)", fontsize=11)
     axC.legend(
-        fontsize=7,
+        fontsize=9,
         loc="lower right",
         frameon=False,
         title="landmark",
-        title_fontsize=7,
+        title_fontsize=9,
     )
-    axC.annotate("1.2 mm", (1.2, 4), fontsize=6.6, color="#111111", ha="left")
-    panel_letter(axC, "C", "One error, calibrated per landmark")
+    panel_letter(axC, "C", "")
 
     _save(fig, "fig5_difficulty")
 
@@ -872,23 +844,6 @@ def fig1_interface():
             color="black",
             zorder=6,
         )
-    legend = (
-        "\u2460  Multiplanar MRI viewer (NiiVue)      "
-        "\u2461  Rater-calibrated result: error + percentile vs raters\n"
-        "\u2462  Difficulty-aware, anatomy-first AI tutor      "
-        "\u2463  Bring-your-own-key + active model      "
-        "\u2464  32-landmark progress + export"
-    )
-    fig.text(
-        0.5,
-        0.05,
-        legend,
-        ha="center",
-        va="center",
-        fontsize=10.5,
-        color="#222222",
-        linespacing=1.9,
-    )
     _save(fig, "fig1_interface")
 
 
@@ -1017,9 +972,11 @@ def fig0_graphical_abstract():
     figure set still regenerates from ``python make_figures.py``.
     """
     import runpy
+    from pathlib import Path as _P
 
-    runpy.run_path("make_ga_assets.py", run_name="__main__")
-    runpy.run_path("make_graphical_abstract.py", run_name="__main__")
+    here = _P(__file__).resolve().parent
+    runpy.run_path(str(here / "make_ga_assets.py"), run_name="__main__")
+    runpy.run_path(str(here / "make_graphical_abstract.py"), run_name="__main__")
 
 
 if __name__ == "__main__":

scripts/make_graphical_abstract.py

@@ -527,50 +527,6 @@ def ic_file(cx, cy, s, color):
         )
 
 
-def ic_chat(cx, cy, s, color):
-    w, h = s * 1.7, s * 1.25
-    rrect(
-        cx - w / 2,
-        cy - h / 2 + 0.12,
-        w,
-        h,
-        fc="white",
-        ec=color,
-        lw=1.6,
-        r=0.14,
-        z=7,
-    )
-    ax.add_patch(
-        Polygon(
-            [
-                (cx - s * 0.2, cy - h / 2 + 0.18),
-                (cx - s * 0.55, cy - h / 2 - 0.14),
-                (cx + s * 0.12, cy - h / 2 + 0.18),
-            ],
-            closed=True,
-            fc="white",
-            ec=color,
-            lw=1.6,
-            zorder=7,
-            joinstyle="round",
-        )
-    )
-    ax.add_patch(
-        Rectangle(
-            (cx - s * 0.5, cy - h / 2 + 0.14),
-            s * 0.7,
-            0.1,
-            fc="white",
-            ec="none",
-            zorder=7.15,
-        )
-    )
-    for dx in (-0.42, 0, 0.42):
-        ax.add_patch(
-            Circle((cx + dx * s, cy + 0.14), s * 0.11, fc=color, zorder=7.2)
-        )
-
-
 def ic_gear(cx, cy, s, color):
     n, outer, inner = 8, s, s * 0.66
     pts = [
@@ -1159,50 +1115,17 @@ def data_row(y, nb, count, label, color):
 )
 
 # ══════════════════════════════════════════════════════════════════════════════
-# THE METRIC — a slim shared header spanning both product cards
-# (the "same mm, two verdicts" proof lives inside the cards themselves)
+# (metric band removed)
 # ══════════════════════════════════════════════════════════════════════════════
-_thesis = "One Euclidean error — two uses"
-_ty = BR_y0 + 0.40
-txt(SEAM, _ty, _thesis, size=11.5, weight="bold", color=SLATE, ha="center")
-txt(
-    SEAM,
-    BR_y0 + 0.17,
-    "the same millimetre, always read against how hard the landmark is",
-    size=8.2,
-    color=MUT,
-    ha="center",
-    style="italic",
-)
-# quiet colour cue mapping the shared metric onto each card, no heavy arrows
-txt(
-    M + 0.30,
-    BR_y0 + 0.13,
-    "taught in context",
-    size=7.8,
-    color=LEARN_C,
-    weight="bold",
-    ha="left",
-)
-txt(
-    XR - 0.30,
-    BR_y0 + 0.13,
-    "audited at scale",
-    size=7.8,
-    color=VALID_C,
-    weight="bold",
-    ha="right",
-)
 
 # ══════════════════════════════════════════════════════════════════════════════
 # 03 — LEARN  (left card)
 # ══════════════════════════════════════════════════════════════════════════════
 LX0 = M
 LX1 = M + CARDW
 card("LEARN", LX0, P_y0, CARDW, H_PROD, LEARN_C)
-ic_chat(LX0 + 0.42, P_y1 - 0.38, 0.18, LEARN_C)
 section(
-    LX0 + 0.78,
+    LX0 + 0.42,
     P_y1 - 0.38,
     "02",
     LEARN_C,
@@ -1345,9 +1268,8 @@ def data_row(y, nb, count, label, color):
 VX0 = M + CARDW + GAP
 VX1 = XR
 card("VALIDATE", VX0, P_y0, CARDW, H_PROD, VALID_C)
-ic_gear(VX0 + 0.42, P_y1 - 0.38, 0.17, VALID_C)
 section(
-    VX0 + 0.78,
+    VX0 + 0.42,
     P_y1 - 0.38,
     "03",
     VALID_C,