scripts/validate_metadata.py

"""
Metadata Validation Script
Checks consistency of S-OH_metadata.csv:
- Age range (15-89 years)
- Gender encoding (0/1)
- ICD-10 code consistency with D_class and D_bin_class mappings
- Temporal consistency: startTimeGases < endTimeGases < durationSec
- Unique patient IDs and no duplicate entries
"""

import pandas as pd
import json
from pathlib import Path

# ============================================
# CONFIGURATION
# ============================================

METADATA_PATH = Path("../metadata.csv")
DATA_DIR = Path("../data")
MANIFEST_PATH = DATA_DIR / "manifest.json"

# ICD-10 to D_class mapping
ICD10_TO_DCLASS = {'Z00': 0, 'E11': 1, 'K29': 2, 'K76': 3, 'B18': 4, 'C34': 5, 'N18': 6, 'J44': 7, 'A15': 8}

# ICD-10 to D_bin_class mapping (for Z00-vs-rest)
ICD10_TO_BIN = {'Z00': 0, 'E11': 1, 'K29': 1, 'K76': 1, 'B18': 1, 'C34': 1, 'N18': 1, 'J44': 1, 'A15': 1}

# Expected columns in metadata
EXPECTED_COLUMNS = [
    'Patient_id', 'Patient_age', 'Patient_gender',
    'Diagnosis', 'D_class', 'D_bin_class',
    'Datetime', 'Week', 'Site'
]


# ============================================
# VALIDATION FUNCTIONS
# ============================================

def validate_age(metadata):
    """Check age range 15-89 years."""
    ages = metadata['Patient_age']
    invalid = (ages < 15) | (ages > 89)
    n_invalid = invalid.sum()
    if n_invalid > 0:
        print(f"❌ Age validation FAILED: {n_invalid} patients outside 15-89 range")
        print(f"   Min: {ages.min()}, Max: {ages.max()}")
        return False
    print(f"✅ Age validation PASSED: {len(ages)} patients, range {ages.min()}-{ages.max()} years")
    return True


def validate_gender(metadata):
    """Check gender encoding (0=female, 1=male)."""
    gender = metadata['Patient_gender']
    invalid = ~gender.isin(['female', 'male'])
    n_invalid = invalid.sum()
    if n_invalid > 0:
        print(f"❌ Gender validation FAILED: {n_invalid} patients with invalid gender encoding")
        print(f"   Unique values: {gender.unique()}")
        return False
    n_female = (gender == 0).sum()
    n_male = (gender == 1).sum()
    print(f"✅ Gender validation PASSED: {n_female} female, {n_male} male")
    return True


def validate_diagnosis_mapping(metadata):
    """Check ICD-10 code consistency with D_class and D_bin_class mappings."""
    errors = []
    
    for idx, row in metadata.iterrows():
        diag = row['Diagnosis']
        d_class = row['D_class']
        d_bin = row['D_bin_class']
        
        expected_class = ICD10_TO_DCLASS.get(diag)
        expected_bin = ICD10_TO_BIN.get(diag)
        
        if expected_class is None:
            errors.append(f"Patient {row['Patient_id']}: unknown diagnosis '{diag}'")
            continue
        
        if d_class != expected_class:
            errors.append(f"Patient {row['Patient_id']}: D_class {d_class} != expected {expected_class}")
        
        if d_bin != expected_bin:
            errors.append(f"Patient {row['Patient_id']}: D_bin_class {d_bin} != expected {expected_bin}")
    
    if errors:
        print(f"❌ Diagnosis mapping validation FAILED: {len(errors)} errors")
        for err in errors[:10]:  # show first 10
            print(f"   {err}")
        if len(errors) > 10:
            print(f"   ... and {len(errors)-10} more")
        return False
    
    print(f"✅ Diagnosis mapping validation PASSED: {len(metadata)} patients")
    return True


def validate_temporal_consistency(metadata, data_dir):
    """Check startTimeGases < endTimeGases < durationSec for each patient."""
    errors = []
    
    # Load manifest to get file paths
    with open(MANIFEST_PATH, 'r') as f:
        manifest = json.load(f)
    
    # Create mapping from patient_id to file path
    patient_file_map = {entry['patient_id']: entry['file'] for entry in manifest['files']}
    
    for patient_id in metadata['Patient_id']:
        if patient_id not in patient_file_map:
            continue
        
        file_path = data_dir / patient_file_map[patient_id]
        if not file_path.exists():
            errors.append(f"Patient {patient_id}: JSON file not found: {file_path}")
            continue
        
        with open(file_path, 'r') as f:
            patient_data = json.load(f)
        
        start_time = patient_data.get('startTimeGases', 0)
        end_time = patient_data.get('endTimeGases', 0)
        duration = patient_data.get('durationSec', 0)
        
        if not (start_time < end_time < duration):
            errors.append(f"Patient {patient_id}: temporal inconsistency: {start_time} < {end_time} < {duration} is False")
    
    if errors:
        print(f"❌ Temporal consistency validation FAILED: {len(errors)} errors")
        for err in errors[:10]:
            print(f"   {err}")
        return False
    
    print(f"✅ Temporal consistency validation PASSED: {len(metadata)} patients")
    return True


def validate_unique_ids(metadata):
    """Check for unique patient IDs and no duplicate entries."""
    patient_ids = metadata['Patient_id']
    n_unique = patient_ids.nunique()
    n_total = len(patient_ids)
    
    if n_unique != n_total:
        duplicates = patient_ids[patient_ids.duplicated()].tolist()
        print(f"❌ Unique IDs validation FAILED: {n_total - n_unique} duplicates found")
        print(f"   Duplicate IDs: {duplicates[:10]}")
        return False
    
    print(f"✅ Unique IDs validation PASSED: {n_unique} unique patients")
    return True


def validate_columns(metadata):
    """Check that all expected columns are present."""
    missing = [col for col in EXPECTED_COLUMNS if col not in metadata.columns]
    if missing:
        print(f"❌ Columns validation FAILED: missing columns: {missing}")
        return False
    print(f"✅ Columns validation PASSED: all {len(EXPECTED_COLUMNS)} columns present")
    return True


def main():
    print("=" * 60)
    print("S-OH METADATA VALIDATION")
    print("=" * 60)
    
    # Load metadata
    print(f"\nLoading metadata from: {METADATA_PATH}")
    metadata = pd.read_csv(METADATA_PATH)
    print(f"Loaded {len(metadata)} records")
    
    all_passed = True
    
    print("\n" + "-" * 40)
    all_passed &= validate_columns(metadata)
    
    print("\n" + "-" * 40)
    all_passed &= validate_unique_ids(metadata)
    
    print("\n" + "-" * 40)
    all_passed &= validate_age(metadata)
    
    print("\n" + "-" * 40)
    all_passed &= validate_gender(metadata)
    
    print("\n" + "-" * 40)
    all_passed &= validate_diagnosis_mapping(metadata)
    
    print("\n" + "-" * 40)
    all_passed &= validate_temporal_consistency(metadata, DATA_DIR)
    
    print("\n" + "=" * 60)
    if all_passed:
        print("✅ ALL CHECKS PASSED")
    else:
        print("❌ SOME CHECKS FAILED")
    print("=" * 60)
    
    return all_passed


if __name__ == "__main__":
    main()