Add inference.py

inference.py

@@ -1,36 +1,66 @@
 """
-Inference script for Autism Spectrum Disorder (ASD) Detector
+Inference script for Simplified ASD Detector (8 features)
 
-Example usage of the trained model for making predictions.
+Example usage for making predictions with the trained model.
 """
 
+import sys
 import json
 import torch
 import joblib
 import pandas as pd
-import numpy as np
 from pathlib import Path
-from model import AutismDetectorNet, load_model
+from model import SimplifiedASDDetector, load_model, FEATURES, SimplePreprocessor
+
+# Fix for unpickling preprocessor saved from different module
+sys.modules['__main__'].SimplePreprocessor = SimplePreprocessor
+
+
+# Original column names (as in the training data)
+ORIGINAL_COLUMN_NAMES = [
+    'Developmental milestones- global delay (G), motor delay (M), cognitive delay (C)',
+    'IQ/DQ',
+    'ICD',
+    'Language disorder Y= present, N=absent',
+    'Language development: delay, normal=N, absent=A',
+    'Dysmorphysm y=present, no=absent',
+    'Behaviour disorder- agressivity, agitation, irascibility',
+    'Neurological Examination; N=normal, text = abnormal; free cell = examination not performed ???'
+]
+
+# Simplified names for user-friendly input
+SIMPLE_NAMES = [
+    'developmental_milestones',
+    'iq_dq',
+    'intellectual_disability',
+    'language_disorder',
+    'language_development',
+    'dysmorphism',
+    'behaviour_disorder',
+    'neurological_exam'
+]
 
 
 class ASDPredictor:
     """
-    Wrapper class for easy ASD prediction.
+    Easy-to-use predictor for ASD detection.
 
     Example:
         >>> predictor = ASDPredictor('.')
-        >>> result = predictor.predict(patient_data)
-        >>> print(result)
+        >>> result = predictor.predict({
+        ...     'developmental_milestones': 'N',
+        ...     'iq_dq': 100,
+        ...     'intellectual_disability': 'N',
+        ...     'language_disorder': 'N',
+        ...     'language_development': 'N',
+        ...     'dysmorphism': 'NO',
+        ...     'behaviour_disorder': 'N',
+        ...     'neurological_exam': 'N'
+        ... })
+        >>> print(result['prediction'])  # 'Healthy' or 'ASD'
     """
 
     def __init__(self, model_dir='.', device='cpu'):
-        """
-        Initialize the predictor.
-
-        Args:
-            model_dir (str): Directory containing model files
-            device (str): Device for inference ('cpu' or 'cuda')
-        """
         model_dir = Path(model_dir)
         self.device = device
 
@@ -44,145 +74,132 @@ class ASDPredictor:
         with open(model_dir / 'config.json', 'r') as f:
             self.config = json.load(f)
 
-        # Load feature info
-        with open(model_dir / 'feature_info.json', 'r') as f:
-            self.feature_info = json.load(f)
-
-        self.feature_columns = self.config['feature_columns']
-
-    def predict(self, data, return_proba=False):
+    def _convert_simple_to_original(self, data):
+        """Convert simplified feature names to original column names."""
+        if isinstance(data, dict):
+            converted = {}
+            for simple, original in zip(SIMPLE_NAMES, ORIGINAL_COLUMN_NAMES):
+                if simple in data:
+                    converted[original] = data[simple]
+                elif original in data:
+                    converted[original] = data[original]
+            return converted
+        return data
+
+    def predict(self, data):
         """
-        Make predictions on patient data.
+        Make prediction on patient data.
 
         Args:
-            data (pd.DataFrame or dict): Patient data with required features
-            return_proba (bool): If True, return probabilities instead of labels
+            data (dict): Patient features using simplified names:
+                - developmental_milestones: N/G/M/C
+                - iq_dq: numeric (e.g., 100)
+                - intellectual_disability: N/F70.0/F71/F72
+                - language_disorder: N/Y
+                - language_development: N/delay/A
+                - dysmorphism: NO/Y
+                - behaviour_disorder: N/Y
+                - neurological_exam: N or abnormal description
 
         Returns:
-            dict: Prediction results including label and probability
+            dict: {
+                'prediction': 'Healthy' or 'ASD',
+                'probability_asd': float,
+                'probability_healthy': float,
+                'confidence': float
+            }
         """
-        # Convert dict to DataFrame if necessary
-        if isinstance(data, dict):
-            data = pd.DataFrame([data])
-
-        # Ensure all required columns are present
-        missing_cols = set(self.feature_columns) - set(data.columns)
-        if missing_cols:
-            raise ValueError(f"Missing required columns: {missing_cols}")
-
-        # Select only required columns in correct order
-        data = data[self.feature_columns]
+        # Convert to original column names
+        converted = self._convert_simple_to_original(data)
+        df = pd.DataFrame([converted])
 
         # Preprocess
-        X = self.preprocessor.transform(data)
+        X = self.preprocessor.transform(df)
         X_tensor = torch.FloatTensor(X).to(self.device)
 
         # Predict
         self.model.eval()
         with torch.no_grad():
-            probabilities = self.model(X_tensor).cpu().numpy().flatten()
-
-        # Format results
-        results = []
-        for prob in probabilities:
-            result = {
-                'prediction': 'ASD' if prob > 0.5 else 'Healthy',
-                'label': 1 if prob > 0.5 else 0,
-                'probability_asd': float(prob),
-                'probability_healthy': float(1 - prob),
-                'confidence': float(max(prob, 1 - prob))
-            }
-            results.append(result)
-
-        return results if len(results) > 1 else results[0]
-
-    def get_required_features(self):
-        """Return list of required feature columns."""
-        return self.feature_columns.copy()
-
-    def get_feature_info(self):
-        """Return detailed feature information."""
-        return self.feature_info.copy()
-
-
-def create_sample_patient():
-    """Create a sample patient record for testing."""
-    return {
-        'Gender': 'M',
-        'Age': 48,
-        'Urban/Rural': 'U',
-        'Pregnancy natural /IVF': 'N',
-        'Single/twins': 'S',
-        'Pregnancy evolution, N=normal, AN=abnormal': 'N',
-        'Birth weeks': 39,
-        'Type of birth': 'N',
-        'Birth Weight g': 3200,
-        'Length at birth  ': 50,
-        'Head circumference at birth  ': 35,
-        'APGAR score': 9,
-        'Postnatal adaptation N=normal, AN=abnormal': 'N',
-        'Developmental milestones- global delay (G), motor delay (M), cognitive delay (C)': 'N',
-        'Other Chronic diseases': 'N',
-        'Infections': 0,
-        'allergies': 0,
-        'Family history psychiatric (P) or Neuro disease (Ne), No=absent': 'NO',
-        'Mother age (years)': 30,
-        'Father age (years)': 32,
-        'IQ/DQ': 100,
-        'ICD': 'N',
-        'Neurological Examination; N=normal, text = abnormal; free cell = examination not performed ???': 'N',
-        'Weight kg': 15,
-        'Height  ': 100,
-        'head circumf  ': 48,
-        'Dysmorphysm y=present, no=absent': 'NO',
-        'malformations Y= present, N=absent': 'N',
-        'Behaviour disorder- agressivity, agitation, irascibility': 'N',
-        'Language development: delay, normal=N, absent=A': 'N',
-        'Language disorder Y= present, N=absent': 'N',
-        'EEG, N=normal, F=focal discharges, G=bilateral discharges': 'N',
-        'MRI structural anomalies of the brain, N=absent, AN=present': 'N'
-    }
+            prob_asd = self.model(X_tensor).cpu().item()
+
+        return {
+            'prediction': 'ASD' if prob_asd > 0.5 else 'Healthy',
+            'label': 1 if prob_asd > 0.5 else 0,
+            'probability_asd': prob_asd,
+            'probability_healthy': 1 - prob_asd,
+            'confidence': max(prob_asd, 1 - prob_asd)
+        }
+
+    @staticmethod
+    def get_feature_info():
+        """Return information about required features."""
+        return FEATURES
 
 
 def main():
-    """Example usage of the ASD predictor."""
+    """Example usage."""
     print("=" * 60)
-    print("ASD Detector - Inference Example")
+    print("ASD Detector - Simplified 8-Feature Model")
     print("=" * 60)
 
-    # Initialize predictor
-    predictor = ASDPredictor(model_dir='.')
-
-    # Create sample patient
-    print("\nSample Patient (Healthy profile):")
-    patient = create_sample_patient()
-    for key, value in list(patient.items())[:10]:
-        print(f"  {key}: {value}")
-    print("  ...")
-
-    # Make prediction
-    print("\nPrediction:")
-    result = predictor.predict(patient)
-    print(f"  Label: {result['prediction']}")
-    print(f"  Probability (ASD): {result['probability_asd']:.4f}")
-    print(f"  Probability (Healthy): {result['probability_healthy']:.4f}")
-    print(f"  Confidence: {result['confidence']:.4f}")
-
-    # Test with ASD-like profile
-    print("\n" + "-" * 60)
-    print("\nSample Patient (ASD-like profile):")
-    patient_asd = create_sample_patient()
-    patient_asd['Developmental milestones- global delay (G), motor delay (M), cognitive delay (C)'] = 'G'
-    patient_asd['Language development: delay, normal=N, absent=A'] = 'delay'
-    patient_asd['IQ/DQ'] = 45
-    patient_asd['Behaviour disorder- agressivity, agitation, irascibility'] = 'Y'
-
-    result_asd = predictor.predict(patient_asd)
-    print(f"  Changed features: Developmental=G, Language=delay, IQ=45, Behaviour=Y")
-    print(f"\nPrediction:")
-    print(f"  Label: {result_asd['prediction']}")
-    print(f"  Probability (ASD): {result_asd['probability_asd']:.4f}")
-    print(f"  Confidence: {result_asd['confidence']:.4f}")
+    predictor = ASDPredictor('.')
+
+    # Example 1: Healthy child profile
+    print("\n--- Example 1: Healthy Child ---")
+    healthy_child = {
+        'developmental_milestones': 'N',  # Normal
+        'iq_dq': 105,                      # Normal IQ
+        'intellectual_disability': 'N',    # None
+        'language_disorder': 'N',          # No
+        'language_development': 'N',       # Normal
+        'dysmorphism': 'NO',               # Absent
+        'behaviour_disorder': 'N',         # No
+        'neurological_exam': 'N'           # Normal
+    }
+
+    print("Input:")
+    for k, v in healthy_child.items():
+        print(f"  {k}: {v}")
+
+    result = predictor.predict(healthy_child)
+    print(f"\nResult: {result['prediction']}")
+    print(f"  Probability ASD: {result['probability_asd']:.2%}")
+    print(f"  Confidence: {result['confidence']:.2%}")
+
+    # Example 2: Child with developmental concerns
+    print("\n--- Example 2: Child with Developmental Concerns ---")
+    concerning_child = {
+        'developmental_milestones': 'G',   # Global delay
+        'iq_dq': 55,                        # Below average
+        'intellectual_disability': 'F70.0', # Mild
+        'language_disorder': 'Y',           # Yes
+        'language_development': 'delay',    # Delayed
+        'dysmorphism': 'NO',                # Absent
+        'behaviour_disorder': 'Y',          # Yes
+        'neurological_exam': 'N'            # Normal
+    }
+
+    print("Input:")
+    for k, v in concerning_child.items():
+        print(f"  {k}: {v}")
+
+    result = predictor.predict(concerning_child)
+    print(f"\nResult: {result['prediction']}")
+    print(f"  Probability ASD: {result['probability_asd']:.2%}")
+    print(f"  Confidence: {result['confidence']:.2%}")
+
+    # Print feature reference
+    print("\n" + "=" * 60)
+    print("FEATURE REFERENCE")
+    print("=" * 60)
+    for name, info in FEATURES.items():
+        print(f"\n{name}:")
+        print(f"  {info['description']}")
+        if isinstance(info['values'], dict):
+            for k, v in info['values'].items():
+                print(f"    '{k}' = {v}")
+        else:
+            print(f"    {info['values']}")
 
 
 if __name__ == '__main__':