Fix: load 30 selected features from JSON instead of guessing from model

src/pwat_estimator.py

@@ -8,6 +8,7 @@ Includes Fitzpatrick-aware debiasing correction factors.
 import numpy as np
 import cv2
 import joblib
+import json
 from dataclasses import dataclass, field
 from typing import Optional
 from pathlib import Path
@@ -22,7 +23,6 @@ ITEM_NAMES = {
     8: "Periulcer Skin",
 }
 
-# Debiasing correction factors (calibrated from 61 DFU images)
 CORRECTION_FACTORS = {
     "I":   {3: 0.0,  4: 0.0,  5: 0.0,  6: 0.0,  7: 0.0,  8: 0.0},
     "II":  {3: 0.0,  4: 0.0,  5: 0.0,  6: 0.0,  7: 0.0,  8: 0.0},
@@ -52,7 +52,6 @@ def extract_features(img_bgr: np.ndarray, ulcer_mask: np.ndarray) -> Optional[di
 
     feats = {}
 
-    # Color features (45)
     hsv = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2HSV).astype(np.float32)
     lab = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2Lab).astype(np.float32)
     rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB).astype(np.float32)
@@ -66,7 +65,6 @@ def extract_features(img_bgr: np.ndarray, ulcer_mask: np.ndarray) -> Optional[di
             feats[f"{cs}_{cn}_p25"] = float(np.percentile(vals, 25))
             feats[f"{cs}_{cn}_p75"] = float(np.percentile(vals, 75))
 
-    # Tissue composition (5)
     h, s, v = hsv[b, 0], hsv[b, 1], hsv[b, 2]
     l_ch = lab[b, 0] * (100 / 255)
     a_ch = lab[b, 1] - 128
@@ -82,7 +80,6 @@ def extract_features(img_bgr: np.ndarray, ulcer_mask: np.ndarray) -> Optional[di
     feats["tissue_necro_pct"] = float(np.sum(necro) / npx * 100)
     feats["tissue_necro_total"] = feats["tissue_eschar_pct"] + feats["tissue_slough_pct"] + feats["tissue_necro_pct"]
 
-    # Morphological features (7)
     mask_u8 = b.astype(np.uint8) if b.dtype == bool else (ulcer_mask > 127).astype(np.uint8)
     cnts, _ = cv2.findContours(mask_u8, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
     if cnts:
@@ -100,7 +97,6 @@ def extract_features(img_bgr: np.ndarray, ulcer_mask: np.ndarray) -> Optional[di
         hull = cv2.convexHull(cnt)
         feats["morph_solidity"] = float(area / (cv2.contourArea(hull) + 1e-8))
 
-    # Texture features (4)
     gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)
     wound_gray = gray[b]
     feats["texture_mean"] = float(np.mean(wound_gray))
@@ -114,7 +110,6 @@ def extract_features(img_bgr: np.ndarray, ulcer_mask: np.ndarray) -> Optional[di
     if np.any(edge_zone):
         feats["edge_gradient"] = float(np.mean(np.abs(cv2.Sobel(gray.astype(np.float32), cv2.CV_32F, 1, 0)[edge_zone])))
 
-    # ROI features (2)
     feats["wound_npx"] = float(npx)
     feats["wound_ratio"] = float(npx / (img_bgr.shape[0] * img_bgr.shape[1]))
 
@@ -126,20 +121,22 @@ class PWATPredictor:
 
     def __init__(self, models_dir: str):
         self.models = {}
-        self.feature_names = {}
         models_path = Path(models_dir)
+
+        # Load the selected feature columns (30 features after variance+correlation filter)
+        features_json = models_path / "selected_features.json"
+        if features_json.exists():
+            with open(features_json) as f:
+                self.selected_features = json.load(f)
+            print(f"PWAT: Loaded {len(self.selected_features)} selected features from JSON")
+        else:
+            self.selected_features = None
+            print("PWAT: WARNING — selected_features.json not found, using all features")
+
         for item in ITEMS:
             pkl = models_path / f"xgb_pwat{item}.pkl"
             if pkl.exists():
-                model = joblib.load(pkl)
-                self.models[item] = model
-                # Extract expected feature names from the trained model
-                try:
-                    names = model.get_booster().feature_names
-                    if names:
-                        self.feature_names[item] = names
-                except Exception:
-                    pass
+                self.models[item] = joblib.load(pkl)
 
     def predict(
         self,
@@ -152,6 +149,14 @@ class PWATPredictor:
         if feats is None:
             return PWATResult(fitzpatrick_type=fitzpatrick_type)
 
+        # Use the exact 30 features from training (order matters)
+        if self.selected_features:
+            cols = self.selected_features
+        else:
+            cols = sorted(feats.keys())
+
+        X = np.array([[feats.get(c, 0.0) for c in cols]])
+
         scores_raw = {}
         scores_adj = {}
         for item in ITEMS:
@@ -160,14 +165,6 @@ class PWATPredictor:
                 scores_adj[item] = 0.0
                 continue
 
-            # Use model's expected feature names if available
-            if item in self.feature_names:
-                cols = self.feature_names[item]
-            else:
-                cols = sorted(feats.keys())
-
-            X = np.array([[feats.get(c, 0.0) for c in cols]])
-
             pred = int(self.models[item].predict(X)[0])
             scores_raw[item] = pred
             factor = CORRECTION_FACTORS.get(fitzpatrick_type, {}).get(item, 0.0)