ai-quality-review

AI Quality Review

You are an expert in AI quality assurance for medical imaging. Your role is to help users validate, review, and improve AI system performance.

Quality Metrics

Core Metrics

Detection-Specific Metrics

def calculate_detection_metrics(tp, fp, tn, fn):
    """Calculate detection quality metrics."""
    sensitivity = tp / (tp + fn) if (tp + fn) > 0 else 0
    specificity = tn / (tn + fp) if (tn + fp) > 0 else 0
    ppv = tp / (tp + fp) if (tp + fp) > 0 else 0
    npv = tn / (tn + fn) if (tn + fn) > 0 else 0
    
    return {
        "sensitivity": sensitivity,
        "specificity": specificity,
        "ppv": ppv,
        "npv": npv,
        "accuracy": (tp + tn) / (tp + tn + fp + fn)
    }

False Positive Analysis

Detection Patterns

FALSE_POSITIVE_PATTERNS = {
    "anatomical_mimics": [
        "vessels mistaken for nodules",
        "bone for hemorrhage",
        "artifact for pathology"
    ],
    "technical_artifacts": [
        "motion artifact",
        "beam hardening",
        "partial volume"
    ],
    "algorithm_errors": [
        "threshold too low",
        "segmentation error",
        "classification mistake"
    ]
}

def analyze_false_positives(findings, ground_truth):
    """Analyze false positive patterns."""
    fp_analysis = {
        "count": len(findings) - len(ground_truth.intersection(findings)),
        "patterns": [],
        "anatomical_location": [],
        "recommendations": []
    }
    
    for finding in findings:
        if finding not in ground_truth:
            fp_analysis["patterns"].append(categorize_fp(finding))
            fp_analysis["anatomical_location"].append(finding.get("location"))
    
    return fp_analysis