File size: 2,835 Bytes
9a1b9a7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
LEVERAGE PAPER RESULTS SUMMARY
================================
Experiment Timestamp: 20251124_152044
WMH Segmentation: Binary vs Three-class Classification Comparison
DATASET INFORMATION:
--------------------
Training Images: 2050
Test Images: 350
Image Size: (256, 256)
Classes: Background (0), Normal WMH (1), Abnormal WMH (2)
METHODOLOGY:
------------
Architecture: Enhanced U-Net with Batch Normalization and Dropout
Loss Functions:
- Scenario 1: weighted_bce
- Scenario 2: weighted_categorical
Training Epochs: 50
Batch Size: 8
Learning Rate: 0.0001
PERFORMANCE RESULTS:
--------------------
| Scenario 1 (Binary) | Scenario 2 (3-class) | Improvement
--------------------|---------------------|----------------------|------------
Accuracy | 0.9852 | 0.9965 | +0.0112
Precision | 0.3340 | 0.7589 | +0.4248
Recall | 0.9682 | 0.7765 | -0.1917
Dice Coefficient | 0.4967 | 0.7676 | +0.2709
IoU Coefficient | 0.3304 | 0.6228 | +0.2924
STATISTICAL SIGNIFICANCE:
-------------------------
DICE COEFFICIENT:
Test: Paired t-test
t-statistic: 9.6244
p-value: 0.0000
Effect Size (Cohen's d): 0.5643
95% Confidence Interval: [0.1353, 0.2051]
Result: SIGNIFICANT improvement
IoU COEFFICIENT:
Test: Paired t-test
t-statistic: 10.1596
p-value: 0.0000
Effect Size (Cohen's d): 0.6481
95% Confidence Interval: [0.1356, 0.2010]
Result: SIGNIFICANT improvement
KEY FINDINGS:
-------------
1. Three-class segmentation shows 56.54% improvement in Dice coefficient
2. Three-class segmentation shows 81.33% improvement in IoU coefficient
3. Dice analysis confirms significant improvement
4. IoU analysis confirms significant improvement
5. Post-processing provided substantial improvements in both scenarios
FILES GENERATED:
----------------
- Models: scenario1_binary_model.h5, scenario2_multiclass_model.h5
- Figures: training_curves.png/.pdf, comparison_visualization.png/.pdf, metrics_comparison.png/.pdf
- Tables: comprehensive_results.csv/.xlsx, latex_table.tex
- Statistics: statistical_analysis.json, statistical_report.txt
- Predictions: All test predictions and ground truth data saved
PUBLICATION READINESS:
----------------------
✓ High-resolution figures (300 DPI, PNG/PDF)
✓ LaTeX-formatted tables
✓ Comprehensive statistical analysis (Dice + IoU)
✓ Post-processing impact analysis
✓ Reproducible results with saved models
✓ Professional documentation
|