Upload 15 files

.gitattributes

@@ -45,3 +45,9 @@ transunet/figures/dice_distribution.png filter=lfs diff=lfs merge=lfs -text
 transunet/figures/dice_iou_distribution.png filter=lfs diff=lfs merge=lfs -text
 transunet/figures/metrics_comparison.png filter=lfs diff=lfs merge=lfs -text
 transunet/figures/training_curves.png filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/comparison_visualization.pdf filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/comparison_visualization.png filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/dice_distribution.png filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/dice_iou_distribution.png filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/metrics_comparison.png filter=lfs diff=lfs merge=lfs -text
+deeplabv3plus/figures/training_curves.png filter=lfs diff=lfs merge=lfs -text

deeplabv3plus/config/experiment_config.json

@@ -0,0 +1,21 @@
+{
+  "timestamp": "20251124_180934",
+  "input_shape": [
+    256,
+    256,
+    1
+  ],
+  "target_size": [
+    256,
+    256
+  ],
+  "epochs": 50,
+  "batch_size": 8,
+  "learning_rate": 0.0001,
+  "validation_split": 0.1,
+  "random_state": 42,
+  "loss_options": {
+    "scenario1": "weighted_bce",
+    "scenario2": "weighted_categorical"
+  }
+}

deeplabv3plus/figures/comparison_visualization.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:327f784dd4c24ff8f6457475578c7aa623501c17be243a593b5767fa748a0f5c
+size 3440927

deeplabv3plus/leverage_summary.txt

@@ -0,0 +1,76 @@
+
+    LEVERAGE PAPER RESULTS SUMMARY
+    ================================
+    Experiment Timestamp: 20251124_180934
+    WMH Segmentation: Binary vs Three-class Classification Comparison
+
+    DATASET INFORMATION:
+    --------------------
+    Training Images: 1044 
+    Test Images: 161
+    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            | 0.9751            | 0.9915             | +0.0164
+    Precision           | 0.2306           | 0.4637            | +0.2331
+    Recall              | 0.9838              | 0.7961               | -0.1876
+    Dice Coefficient    | 0.3736                | 0.5861                 | +0.2125
+    IoU Coefficient     | 0.2297                 | 0.4145                  | +0.1848
+
+    STATISTICAL SIGNIFICANCE:
+    -------------------------
+    DICE COEFFICIENT:
+    Test: Paired t-test
+    t-statistic: 9.1289
+    p-value: 0.0000
+    Effect Size (Cohen's d): 0.5655
+    95% Confidence Interval: [0.1278, 0.1983]
+    Result: SIGNIFICANT improvement
+
+    IoU COEFFICIENT:
+    Test: Paired t-test
+    t-statistic: 9.2000
+    p-value: 0.0000
+    Effect Size (Cohen's d): 0.6282
+    95% Confidence Interval: [0.1177, 0.1821]
+    Result: SIGNIFICANT improvement
+
+    KEY FINDINGS:
+    -------------
+    1. Three-class segmentation shows 72.03% improvement in Dice coefficient
+    2. Three-class segmentation shows 99.70% 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
+    

deeplabv3plus/tables/comprehensive_results.csv

@@ -0,0 +1,4 @@
+Scenario,Accuracy,Precision,Recall,Specificity,Dice,IoU,HD95,ASSD
+Binary Classification (Processed),0.9751426744164887,0.23057465300298352,0.9837612105508107,0.9998735813397832,0.3735875824298393,0.22970040142536163,,
+Three-class Classification (Processed),0.9915259698903338,0.4636951727876803,0.7961283505867851,0.9984437529711327,0.5860517227030525,0.4144788980484009,,
+Statistical Analysis,Dice p=0.0000,Dice t=9.1289,Dice Δ=0.1630,Dice ES=0.5655,IoU p=0.0000,IoU Δ=0.1499,HD95 Δ=0.4319px,ASSD Δ=-1.5085px

deeplabv3plus/tables/latex_table.tex

@@ -0,0 +1,12 @@
+\begin{table}
+\caption{Performance comparison between binary and three-class segmentation approaches}
+\label{tab:performance_comparison}
+\begin{tabular}{lllllllll}
+\toprule
+Scenario & Accuracy & Precision & Recall & Specificity & Dice & IoU & HD95 & ASSD \\
+\midrule
+Binary Classification (Processed) & 0.9751 & 0.2306 & 0.9838 & 0.9999 & 0.3736 & 0.2297 & NaN & NaN \\
+Three-class Classification (Processed) & 0.9915 & 0.4637 & 0.7961 & 0.9984 & 0.5861 & 0.4145 & NaN & NaN \\
+\bottomrule
+\end{tabular}
+\end{table}