from typing import Dict, Any

def generated_structured_explanation(features: Dict[str, float], prob: float) -> Dict[str, Any]:
    if prob > 0.80:
        confidence = "High"
        label = "AI"
    elif prob > 0.60:
        confidence = "Moderate"
        label = "Suspicious"
    elif prob > 0.50:
        confidence = "Low"
        label = "Suspicious"
    else:
        confidence = "High" if prob < 0.20 else "Moderate"
        label = "Real"

    contributions = {}
    region_analysis = []

    freq = features.get("frequency_score", 0.0)
    cnn = features.get("cnn_score", 0.0)
    geometry_score = features.get("geometry_score")
    geometry_message = features.get("geometry_message")
    
    if freq > 1.5:
        contributions["frequency"] = freq
        region_analysis.append({
            "region": "Global High-Frequency", 
            "reason": f"Frequency variance score of {freq:.2f} strongly indicates synthetic noise patterns (Target > 1.5)."
        })
        
    if cnn > 1.0:
        contributions["cnn"] = cnn
        region_analysis.append({
            "region": "CNN Activation Areas", 
            "reason": f"CNN artifact score of {cnn:.2f} indicates structural anomalies in bounding edges (Target > 1.0)."
        })
        
    if geometry_score is not None and geometry_score < 50:
        contributions["geometry"] = geometry_score
        region_analysis.append({
            "region": "Perspective & Structural Lines",
            "reason": f"Perspective consistency score of {geometry_score:.1f}/100. {geometry_message}"
        })
    elif geometry_score is not None and geometry_score >= 75:
        region_analysis.append({
            "region": "Perspective & Structural Lines",
            "reason": f"Perspective consistency score of {geometry_score:.1f}/100. {geometry_message}"
        })
        
    if not region_analysis:
        region_analysis.append({
            "region": "Image-Wide",
            "reason": "Measurements align with natural photographic noise distribution baselines."
        })
    return {
        "label": label,
        "probability": float(prob),
        "confidence_level": confidence,
        "feature_contributions": contributions,
        "region_analysis": region_analysis
    }