Update inference.py

inference.py

@@ -14,13 +14,11 @@ print(f"Working directory: {os.getcwd()}")
 print(f"Files present: {os.listdir('.')}")
 
 # ---------------------------------------------------------------------------
-# Compatibility patch — inject _RemainderColsList if the installed sklearn
-# version does not have it (added in sklearn 1.4+).
+# Compatibility patch
 # ---------------------------------------------------------------------------
 import sklearn.compose._column_transformer as _ct
 if not hasattr(_ct, "_RemainderColsList"):
     class _RemainderColsList(list):
-        """Minimal shim for sklearn._RemainderColsList (missing in this env)."""
         def __init__(self, lst=None, future_dtype=None):
             super().__init__(lst or [])
             self.future_dtype = future_dtype
@@ -70,11 +68,10 @@ DEFAULT_N_BOOT_CI   = 500
 
 
 # ---------------------------------------------------------------------------
-# Model loading — skops
+# Model loading
 # ---------------------------------------------------------------------------
 
 def _load_skops_model(fname):
-    """Load a skops model file. Raises RuntimeError on failure (no sys.exit)."""
     if not os.path.exists(fname):
         raise RuntimeError(f"Model file not found: {fname}")
     try:
@@ -106,26 +103,19 @@ consensus_thresholds  = {
     for o, d in classification_model_data.items()
 }
 
-# Calibrators — isotonic only; supports both old and new key names
 calibrators = {}
 for _o, _d in classification_model_data.items():
     _cal      = None
     _cal_type = _d.get("calibrator_type", None)
-
     if "calibrator" in _d and _d["calibrator"] is not None:
         if _cal_type is None or _cal_type == "isotonic":
             _cal = _d["calibrator"]
         else:
-            print(
-                f"  Warning: outcome '{_o}' has calibrator_type='{_cal_type}'. "
-                "Skipping non-isotonic calibrator (isotonic-only policy)."
-            )
+            print(f"  Warning: outcome '{_o}' has calibrator_type='{_cal_type}'. Skipping.")
     elif "isotonic_calibrator" in _d and _d["isotonic_calibrator"] is not None:
         _cal = _d["isotonic_calibrator"]
-
     calibrators[_o] = _cal
 
-# Alias expected by app.py
 isotonic_calibrators = calibrators
 
 oof_probs_calibrated = {
@@ -255,14 +245,11 @@ def bootstrap_ci_from_oof(
 def _calibrate_point(outcome: str, raw_prob: float, use_calibration: bool) -> float:
     beta   = betas[outcome]
     p_beta = float(calibrate_probabilities_undersampling([raw_prob], beta)[0])
-
     if not use_calibration:
         return p_beta
-
     cal = calibrators.get(outcome)
     if cal is None:
         return p_beta
-
     return float(cal.transform([p_beta])[0])
 
 
@@ -317,7 +304,6 @@ def predict_all_outcomes(
     if "DEAD" in probs:
         p_dead      = probs["DEAD"]
         probs["OS"] = float(1.0 - p_dead)
-
         dead_lo, dead_hi = intervals["DEAD"]
         intervals["OS"]  = (
             float(np.clip(1.0 - dead_hi, 0, 1)),
@@ -355,9 +341,10 @@ def predict_all_outcomes(
                 )
                 for _ in range(n_boot_ci)
             ])
-            efs_lo           = float(np.percentile(efs_boot, 2.5))
-            efs_hi           = float(np.percentile(efs_boot, 97.5))
-            intervals["EFS"] = (efs_lo, efs_hi)
+            intervals["EFS"] = (
+                float(np.percentile(efs_boot, 2.5)),
+                float(np.percentile(efs_boot, 97.5)),
+            )
         else:
             intervals["EFS"] = (probs["EFS"], probs["EFS"])
 
@@ -375,22 +362,18 @@ def predict_with_comparison(user_inputs, n_boot_ci: int = DEFAULT_N_BOOT_CI):
 # ---------------------------------------------------------------------------
 
 def _get_shap_values_for_model_outcome(user_inputs, model_outcome, invert, X_proc):
-    """Return per-model SHAP values (shape: n_models x n_processed_features)."""
     all_model_shap_vals = []
     for rf_model in classification_models[model_outcome]:
         explainer = shap.TreeExplainer(rf_model, model_output="probability", data=shap_background)
         shap_vals = explainer.shap_values(X_proc)
-
         if isinstance(shap_vals, list):
             shap_vals = shap_vals[1]
         elif shap_vals.ndim == 3 and shap_vals.shape[2] == 2:
             shap_vals = shap_vals[:, :, 1]
-
         sv = shap_vals[0]
         if invert:
             sv = -sv
         all_model_shap_vals.append(sv)
-
     return np.array(all_model_shap_vals)
 
 
@@ -518,9 +501,85 @@ def create_all_shap_plots(user_inputs, max_display=10):
 
 
 # ---------------------------------------------------------------------------
-# Icon array
+# Icon array  — proper stick figures
 # ---------------------------------------------------------------------------
 
+def _stick_figure_shapes(cx, cy, color, s=0.42):
+    """
+    Draw a classic stick figure centred at (cx, cy) with scale s.
+
+    Parts:
+      - head   : circle
+      - spine  : vertical line from neck to hips
+      - arms   : angled lines left/right from mid-spine
+      - legs   : angled lines left/right from hips
+    """
+    shapes = []
+    lw = dict(color=color, width=max(1, s * 4))   # line width scales with size
+
+    # ── head ──────────────────────────────────────────────────────────────
+    hr = s * 0.22                      # head radius
+    hy = cy + s * 0.60                 # head centre y
+    shapes.append(dict(
+        type="circle", xref="x", yref="y",
+        x0=cx - hr, y0=hy - hr,
+        x1=cx + hr, y1=hy + hr,
+        fillcolor=color,
+        line=dict(color=color, width=0),
+    ))
+
+    # key y levels
+    neck_y  = cy + s * 0.35            # base of head / top of spine
+    hip_y   = cy - s * 0.20            # bottom of spine / top of legs
+    arm_y   = cy + s * 0.15            # where arms branch from spine
+    foot_y  = cy - s * 0.60            # feet
+
+    # ── spine ─────────────────────────────────────────────────────────────
+    shapes.append(dict(
+        type="line", xref="x", yref="y",
+        x0=cx, y0=neck_y,
+        x1=cx, y1=hip_y,
+        line=lw,
+    ))
+
+    # ── arms (angled outward) ──────────────────────────────────────────────
+    arm_dx = s * 0.35
+    arm_dy = s * 0.18
+    # left arm
+    shapes.append(dict(
+        type="line", xref="x", yref="y",
+        x0=cx,          y0=arm_y,
+        x1=cx - arm_dx, y1=arm_y - arm_dy,
+        line=lw,
+    ))
+    # right arm
+    shapes.append(dict(
+        type="line", xref="x", yref="y",
+        x0=cx,          y0=arm_y,
+        x1=cx + arm_dx, y1=arm_y - arm_dy,
+        line=lw,
+    ))
+
+    # ── legs (angled outward) ──────────────────────────────────────────────
+    leg_dx = s * 0.28
+    # left leg
+    shapes.append(dict(
+        type="line", xref="x", yref="y",
+        x0=cx,         y0=hip_y,
+        x1=cx - leg_dx, y1=foot_y,
+        line=lw,
+    ))
+    # right leg
+    shapes.append(dict(
+        type="line", xref="x", yref="y",
+        x0=cx,         y0=hip_y,
+        x1=cx + leg_dx, y1=foot_y,
+        line=lw,
+    ))
+
+    return shapes
+
+
 def icon_array(probability, outcome):
     outcome_labels = {
         "DEAD":     ("Death",           "Overall Survival"),
@@ -537,51 +596,15 @@ def icon_array(probability, outcome):
     n_no_event = n_total - n_event
     cols, rows = 10, 10
 
-    shapes   = []
-    icon_idx = 0
-    sx, sy   = 0.38, 0.38   # icon half-size within each unit cell
+    all_shapes = []
+    icon_idx   = 0
 
-    for row in range(rows):          # row 0 = top
-        for col in range(cols):      # col 0 = left
-            color = "#ff6b6b" if icon_idx < n_event else "#4ecdc4"
+    for row in range(rows):        # row 0 = top
+        for col in range(cols):    # col 0 = left
+            color = "#e05555" if icon_idx < n_event else "#3bbfad"
             cx = col
-            cy = rows - 1 - row      # invert so row 0 renders at the top
-
-            # ── head (circle) ──────────────────────────────────────────
-            hr = sy * 0.22
-            hx = cx
-            hy = cy + sy * 0.65
-            shapes.append(dict(
-                type="circle", xref="x", yref="y",
-                x0=hx - hr, y0=hy - hr,
-                x1=hx + hr, y1=hy + hr,
-                fillcolor=color, line=dict(color=color, width=0),
-            ))
-
-            # ── body (symmetric trapezoid: shoulders → waist → feet) ───
-            tx0 = cx - sx * 0.32
-            tx1 = cx + sx * 0.32
-            wx0 = cx - sx * 0.20
-            wx1 = cx + sx * 0.20
-            bx0 = cx - sx * 0.32
-            bx1 = cx + sx * 0.32
-
-            ty_top    = cy + sy * 0.38   # shoulder line
-            ty_waist  = cy + sy * 0.00   # waist line
-            ty_bottom = cy - sy * 0.42   # feet line
-
-            shapes.append(dict(
-                type="path", xref="x", yref="y",
-                path=(
-                    f"M {tx0},{ty_top} "
-                    f"L {tx1},{ty_top} "
-                    f"L {wx1},{ty_waist} "
-                    f"L {bx1},{ty_bottom} "
-                    f"L {bx0},{ty_bottom} "
-                    f"L {wx0},{ty_waist} Z"
-                ),
-                fillcolor=color, line=dict(color=color, width=0),
-            ))
+            cy = rows - 1 - row    # invert so row 0 is at top
+            all_shapes.extend(_stick_figure_shapes(cx, cy, color, s=0.38))
             icon_idx += 1
 
     fig = go.Figure()
@@ -589,27 +612,27 @@ def icon_array(probability, outcome):
         title=dict(
             text=(
                 f"<b>{OUTCOME_DESCRIPTIONS.get(outcome, outcome)}</b><br>"
-                f"<span style='font-size:12px;color:#ff6b6b'>■ {event_label}: {n_event}%</span>"
+                f"<span style='font-size:12px;color:#e05555'>■ {event_label}: {n_event}%</span>"
                 f"&nbsp;&nbsp;"
-                f"<span style='font-size:12px;color:#4ecdc4'>■ {no_event_label}: {n_no_event}%</span>"
+                f"<span style='font-size:12px;color:#3bbfad'>■ {no_event_label}: {n_no_event}%</span>"
             ),
             x=0.5, xanchor="center",
             font=dict(size=14, color="black"),
         ),
-        shapes=shapes,
+        shapes=all_shapes,
         xaxis=dict(
-            range=[-0.65, cols - 0.35],
+            range=[-0.7, cols - 0.3],
             showgrid=False, zeroline=False, showticklabels=False,
             fixedrange=True,
         ),
         yaxis=dict(
-            range=[-0.65, rows - 0.35],
+            range=[-0.8, rows - 0.2],
             showgrid=False, zeroline=False, showticklabels=False,
             fixedrange=True,
         ),
-        height=420,
-        width=400,
-        margin=dict(l=10, r=10, t=80, b=10),
+        height=440,
+        width=420,
+        margin=dict(l=10, r=10, t=85, b=10),
         plot_bgcolor="white",
         paper_bgcolor="white",
     )