File size: 7,156 Bytes
eded815
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
---
language:
- en
license: mit
tags:
- medical-imaging
- chest-xray
- pneumonia-detection
- state-space-models
- mamba
- sssm
- pediatric
- physics-informed-neural-networks
pipeline_tag: image-classification
datasets:
- paultimothymooney/chest-xray-pneumonia
metrics:
- accuracy
- f1
- precision
- recall
- roc_auc
model-index:
- name: BioHCL-SSSM
  results:
  - task:
      type: image-classification
      name: Pediatric Pneumonia Classification
    dataset:
      name: Guangzhou Pediatric Chest X-Ray Cohort (Cell 2018)
      type: paultimothymooney/chest-xray-pneumonia
    metrics:
    - type: accuracy
      value: 99.37
      name: 10-Fold Cross-Validation Accuracy
    - type: f1
      value: 0.9957
      name: 10-Fold Cross-Validation F1
    - type: roc_auc
      value: 0.9929
      name: 10-Fold Cross-Validation AUC
---

# BioHCL-SSSM: Complete Medical Imaging Workspace & Model Hub

**Author:** unknownBrain  
**Project:** BioHCL-SSSM: Hierarchical Convolutional State-Space Hybrid Architecture for Pediatric Pneumonia Diagnosis  
**Dataset:** Guangzhou Women and Children’s Medical Center Pediatric Chest X-Ray Cohort (Kermany et al., *Cell* 2018; 5,856 images, 624 official held-out test split)

---

## πŸ“Œ Complete Repository Contents

This public repository contains the complete end-to-end workspace for the BioHCL-SSSM project, including all pre-trained checkpoints, raw experimental sweeps, modular source code, deliverable packages, evaluation figures, and peer-reviewed reference literature:

```
BioHCL-SSSM-Complete-Workspace/
β”‚
β”œβ”€β”€ deliverables/                              # πŸ“ OFFICIAL SUPERVISOR DELIVERABLE PACKAGES
β”‚   β”œβ”€β”€ 1_sota_benchmarks_master.csv           # 15-Model SOTA Benchmark (Literature baselines + Ours)
β”‚   β”œβ”€β”€ 2_table1_component_ablation.csv        # Table 1: Macro Component Ablation (SSSM vs Attention vs CNN)
β”‚   β”œβ”€β”€ 3_table2_layer_hyperparameter_deduction.csv # Table 2: Proposed SSSM Layer Hyperparameter Deduction
β”‚   β”œβ”€β”€ 4_table3_loss_function_weighting.csv   # Table 3: Loss Function & Class Weighting Dynamics
β”‚   β”œβ”€β”€ 5_10_fold_cross_validation_metrics.csv # Approved 10-Fold CV metrics (Folds 1 to 10 + Mean/Std)
β”‚   β”œβ”€β”€ remaining_ablations_plan.md            # Comprehensive scientific audit & execution plan
β”‚   β”œβ”€β”€ tathagat_sir_revision_report.md        # Full revision report formatted for Tathagat Sir
β”‚   β”œβ”€β”€ v2/                                    # Clean v2 supervisor release package
β”‚   β”œβ”€β”€ raw_sweeps/                            # Raw 64-run ablation search & hardware profiling logs
β”‚   └── figures/                               # Publication-ready ROC, PR, and Confusion Matrix plots
β”‚
β”œβ”€β”€ medical imaging/                           # πŸ“ FULL ORIGINAL TRAINING CODE & WEIGHTS
β”‚   β”œβ”€β”€ best_student_cv.pt                     # Best Model Checkpoint (Fold 2: 99.81% Val Acc, 0.9987 F1)
β”‚   β”œβ”€β”€ model_fold_1.pt … model_fold_10.pt     # All 10 trained cross-validation model weights (54 MB each)
β”‚   β”œβ”€β”€ ablation.csv                           # Raw 64 empirical ablation sweep runs
β”‚   β”œβ”€β”€ extracted_code.py                      # Original full pipeline code extracted from Colab
β”‚   β”œβ”€β”€ build_benchmark.py                     # Benchmark runner
β”‚   β”œβ”€β”€ build_ablations.py                     # Ablation engine runner
β”‚   β”œβ”€β”€ build_eval.py                          # Test evaluation engine
β”‚   β”œβ”€β”€ build_xai.py                           # Explainability generator (Attention, GradCAM, LIME, SHAP)
β”‚   └── *.ipynb / *.ipynb.txt                  # Full Colab training notebooks
β”‚
β”œβ”€β”€ hf_repo_staging/                           # πŸ“ CLEAN MODULARIZED SOURCE & ARTIFACTS
β”‚   β”œβ”€β”€ src/                                   # Clean modular Python package (models, dataset, loss, eval)
β”‚   β”œβ”€β”€ checkpoints/                           # Reorganized weights (cross_validation/ and ablations/)
β”‚   β”œβ”€β”€ results/                               # CSV deliverables
β”‚   └── figures/                               # Complete high-res figures & XAI heatmaps
β”‚
β”œβ”€β”€ papers to refer/                           # πŸ“ 12 PEER-REVIEWED REFERENCE PAPERS
β”‚   β”œβ”€β”€ IBMED-S-26-01139.pdf                   # Bio-HCLSSSM Reference Paper (Computers in Biology and Medicine)
β”‚   β”œβ”€β”€ CMPB-D-26-01992_R1_reviewer.pdf        # Computer Methods and Programs in Biomedicine
β”‚   β”œβ”€β”€ EAAI-26-17234_reviewer.pdf             # Engineering Applications of Artificial Intelligence
β”‚   β”œβ”€β”€ TCIV-2026-0179.pdf                     # Computer Methods in Biomechanics and Biomedical Engineering
β”‚   β”œβ”€β”€ IJIG-D-26-00102_R1_reviewer.pdf        # International Journal of Image and Graphics
β”‚   β”œβ”€β”€ JBHI-04026-2026_Proof_hi.pdf           # IEEE Journal of Biomedical and Health Informatics
β”‚   └── ... (6 additional reviewer proofs)
β”‚
β”œβ”€β”€ run_colab_remaining_ablations.py           # πŸš€ STANDALONE RUNNER FOR REMAINING EXPERIMENTS
β”‚   # Executes the 4 remaining core runs (lambda_bio in [0.2, 0.5, 1.0] and MHSA Attention baseline)
β”‚   # Takes ~10 mins on GPU and logs directly to CSV
β”‚
β”œβ”€β”€ 10_fold_metrics.csv                        # Exact 10-fold cross-validation metrics
β”œβ”€β”€ benchmark_master.csv                       # Consolidated SOTA benchmarks
β”œβ”€β”€ remaining_ablations_plan.md                # Scientific audit & plan
└── tathagat_sir_revision_report.md            # Comprehensive revision report
```

---

## πŸ† Key Performance Highlights

1. **Approved 10-Fold Cross-Validation:**
   * **Mean Accuracy:** **99.37% Β± 0.33%**
   * **Mean F1-Score:** **0.9957 Β± 0.0022**
   * **Mean ROC-AUC:** **0.9929 Β± 0.0039**
   * **Pooled Errors:** Only 33 errors across 5,232 images (3,862 TP, 1,337 TN, 12 FP, 21 FN).
   * **Peak Fold:** **Fold 2** achieved **99.81% accuracy, 0.9987 F1, 100% recall, 99.26% specificity** (`best_student_cv.pt`).

2. **Official Held-Out Test Set (624 Images) Benchmark:**
   * **BioHCL-SSSM:** **96.47% Accuracy, 97.22% F1, 99.37% ROC-AUC** with only **14.01M parameters and 8.56 GFLOPs**.
   * Outperforms standard Vision Transformers (**ViT-Base:** 86.57M params, 92.47% Acc) with **83.8% fewer parameters**.
   * Matches/exceeds Hierarchical Shifted-Window Transformers (**Swin-T:** 27.52M params, 95.99% Acc) with **49.1% fewer parameters**.

---

## πŸš€ Quick Execution on New GPU (i9 + RTX 3080 Ti or Colab)

To execute the remaining 4 core ablation runs:

```bash
# 1. Clone this repository
git clone https://huggingface.co/unknownBrain/BioHCL-SSSM-Complete-Workspace
cd BioHCL-SSSM-Complete-Workspace

# 2. Install dependencies
pip install torch torchvision timm albumentations huggingface_hub scikit-learn pandas pillow

# 3. Run the automated ablation suite
python run_colab_remaining_ablations.py
```

---

## πŸ“œ License & Citation

Licensed under the **MIT License**.  
Guangzhou Women and Children’s Medical Center Chest X-Ray Cohort (Kermany et al., *Cell* 2018).