diagnose_model_outputs.py

#!/usr/bin/env python3
"""
Diagnostic Script for ECG-FM Model Outputs
Examines actual model outputs to understand clinical analysis issues
"""

import pandas as pd
import requests
import json
import time
import os

# Configuration
API_URL = "https://mystic-cbk-ecg-fm-api.hf.space"
ECG_FILE = "../ecg_uploads_greenwich/ecg_98408931-6f8e-47cc-954a-ba0c058a0f3d.csv"

def diagnose_model_outputs():
    """Diagnose what the models are actually outputting"""
    print("🔍 DIAGNOSING ECG-FM MODEL OUTPUTS")
    print("=" * 60)
    print(f"🌐 API URL: {API_URL}")
    print(f"📁 ECG File: {ECG_FILE}")
    print()
    
    try:
        # 1. Load ECG data
        print("📁 Loading ECG data...")
        if not os.path.exists(ECG_FILE):
            print(f"❌ ECG file not found: {ECG_FILE}")
            return
        
        df = pd.read_csv(ECG_FILE)
        signal = [df[col].tolist() for col in df.columns]
        
        payload = {
            "signal": signal,
            "fs": 500,
            "lead_names": ["I", "II", "III", "aVR", "aVL", "aVF", "V1", "V2", "V3", "V4", "V5", "V6"]
        }
        
        print(f"✅ Loaded ECG: {len(signal)} leads, {len(signal[0])} samples")
        
        # 2. Test feature extraction (pretrained model)
        print("\n🧬 Testing Feature Extraction (Pretrained Model)...")
        print("   This should show what the pretrained model outputs")
        
        feature_response = requests.post(
            f"{API_URL}/extract_features",
            json=payload,
            timeout=120
        )
        
        if feature_response.status_code == 200:
            feature_data = feature_response.json()
            print("✅ Feature extraction successful!")
            print(f"   Features count: {len(feature_data.get('features', []))}")
            print(f"   Input shape: {feature_data.get('input_shape', 'Unknown')}")
            print(f"   Model type: {feature_data.get('model_type', 'Unknown')}")
            
            # Show physiological parameters
            phys_params = feature_data.get('physiological_parameters', {})
            if phys_params:
                print(f"   Physiological parameters: {len(phys_params)}")
                for key, value in phys_params.items():
                    print(f"     {key}: {value}")
        else:
            print(f"❌ Feature extraction failed: {feature_response.status_code}")
            print(f"   Response: {feature_response.text}")
            return
        
        # 3. Test full analysis (both models)
        print("\n🏥 Testing Full Analysis (Both Models)...")
        print("   This should show what both models output together")
        
        analysis_response = requests.post(
            f"{API_URL}/analyze",
            json=payload,
            timeout=180
        )
        
        if analysis_response.status_code == 200:
            analysis_data = analysis_response.json()
            print("✅ Full analysis successful!")
            
            # Examine clinical analysis
            clinical = analysis_data.get('clinical_analysis', {})
            print(f"\n📊 Clinical Analysis Details:")
            print(f"   Rhythm: {clinical.get('rhythm', 'Unknown')}")
            print(f"   Heart Rate: {clinical.get('heart_rate', 'Unknown')} BPM")
            print(f"   QRS Duration: {clinical.get('qrs_duration', 'Unknown')} ms")
            print(f"   QT Interval: {clinical.get('qt_interval', 'Unknown')} ms")
            print(f"   PR Interval: {clinical.get('pr_interval', 'Unknown')} ms")
            print(f"   Axis Deviation: {clinical.get('axis_deviation', 'Unknown')}")
            print(f"   Confidence: {clinical.get('confidence', 'Unknown')}")
            print(f"   Method: {clinical.get('method', 'Unknown')}")
            
            # Check for probabilities
            if 'probabilities' in clinical:
                probs = clinical['probabilities']
                print(f"   Probabilities count: {len(probs)}")
                if len(probs) > 0:
                    print(f"   First 5 probabilities: {probs[:5]}")
                    print(f"   Max probability: {max(probs):.4f}")
                    print(f"   Min probability: {min(probs):.4f}")
            
            # Check for label probabilities
            if 'label_probabilities' in clinical:
                label_probs = clinical['label_probabilities']
                print(f"   Label probabilities: {len(label_probs)}")
                if label_probs:
                    print(f"   Sample labels: {list(label_probs.keys())[:5]}")
            
            # Check for abnormalities
            abnormalities = clinical.get('abnormalities', [])
            print(f"   Abnormalities: {abnormalities}")
            
            # Examine physiological parameters
            phys_params = clinical.get('physiological_parameters', {})
            if phys_params:
                print(f"\n📊 Physiological Parameters (from clinical analysis):")
                for key, value in phys_params.items():
                    print(f"   {key}: {value}")
            
            # Examine features
            features = analysis_data.get('features', [])
            print(f"\n📊 Features:")
            print(f"   Count: {len(features)}")
            if len(features) > 0:
                print(f"   First 5 values: {features[:5]}")
                print(f"   Last 5 values: {features[-5:]}")
            
            # Examine signal quality
            signal_quality = analysis_data.get('signal_quality', 'Unknown')
            print(f"\n📊 Signal Quality: {signal_quality}")
            
            # Examine processing time
            processing_time = analysis_data.get('processing_time', 'Unknown')
            print(f"⏱️  Processing Time: {processing_time}s")
            
        else:
            print(f"❌ Full analysis failed: {analysis_response.status_code}")
            print(f"   Response: {analysis_response.text}")
            return
        
        # 4. Summary and diagnosis
        print("\n" + "=" * 60)
        print("🔍 DIAGNOSIS SUMMARY")
        print("=" * 60)
        
        if clinical.get('method') == 'clinical_predictions':
            print("✅ Clinical analysis method: clinical_predictions")
            print("   This means the finetuned model is working")
        else:
            print("❌ Clinical analysis method: NOT clinical_predictions")
            print("   This means the finetuned model is not producing proper outputs")
        
        if clinical.get('probabilities'):
            print("✅ Probabilities are available")
            print(f"   Count: {len(clinical['probabilities'])}")
        else:
            print("❌ No probabilities available")
            print("   This explains why clinical analysis is failing")
        
        if clinical.get('rhythm') != 'Unable to determine':
            print("✅ Rhythm detection working")
        else:
            print("❌ Rhythm detection failing")
            print("   This suggests clinical model output issues")
        
        print(f"\n🎯 RECOMMENDED ACTIONS:")
        print(f"   1. Check if finetuned model is producing label_logits")
        print(f"   2. Verify model output format matches expectations")
        print(f"   3. Debug clinical_analysis.py logic")
        print(f"   4. Test with simpler ECG data")
        
    except Exception as e:
        print(f"❌ Diagnosis failed with error: {e}")
        import traceback
        traceback.print_exc()

if __name__ == "__main__":
    diagnose_model_outputs()