File size: 1,991 Bytes
70fd054
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d4a1e63
70fd054
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# DermAssist – ResNet50 Ensemble

## Model Overview

DermAssist is a dermoscopic skin lesion malignancy risk estimation model built using a 3-model ResNet50 ensemble.

The system outputs:

- Malignancy probability
- Confidence score
- Uncertainty estimation (ensemble disagreement)

This model is designed for dermoscopic image triage workflows.

---

## Architecture

Backbone:
ResNet50 (ImageNet pretrained)

Classifier Head:
Linear → ReLU → Dropout (0.5) → Linear → 1 output logit

Training Details:
- Optimizer: AdamW (lr = 1e-4, weight_decay = 1e-4)
- Loss: BCEWithLogitsLoss (class-weighted)
- Early Stopping: Validation AUC
- Full fine-tuning applied

Ensemble Strategy:
Three independently trained models.  
Final probability = mean prediction.  
Uncertainty = standard deviation across models.

---

## Performance

### Internal Validation (HAM10000)

| Metric | Value |
|--------|-------|
| AUC-ROC | 0.937 |
| Accuracy | 0.86 |
| Malignant Recall | > 0.90 |
| Malignant Precision | ~0.52 |
| Benign Precision | ~0.96 |

---

### External Validation (ISIC 2019 – Filtered Classes)

| Metric | Value |
|--------|-------|
| AUC-ROC | ~0.72 |
| Accuracy | ~0.57 |

Observed performance variation reflects dataset distribution differences.

---

## Intended Use

This model is intended for:

- Dermoscopic image triage
- Educational experimentation
- Ensemble uncertainty analysis
- Explainability demonstration

This model is not intended for autonomous medical diagnosis.

---

## Limitations

- Trained on dermoscopic images only  
- Binary mapping reduces lesion taxonomy complexity  
- Sensitive to dataset distribution shift  
- Requires calibrated threshold tuning  

---

## Files in Repository

- resnet50_model_1.pth  
- resnet50_model_2.pth  
- resnet50_model_3.pth  

All three models are required for ensemble inference.

---

## Ethical Considerations

AI-based medical systems must be used responsibly.  
This model does not replace dermatologist evaluation.