Upload folder using huggingface_hub

README.md

@@ -1,73 +1,24 @@
 ---
 license: cc-by-4.0
 tags:
-- medical-imaging
-- ultrasound
-- thyroid
-- classification
-- efficientnet
-- ml-intern
+  - medical-imaging
+  - ultrasound
+  - thyroid
+  - classification
+  - efficientnet
 datasets:
-- Johnyquest7/TN5000-thyroid-nodule-classification
+  - Johnyquest7/TN5000-thyroid-nodule-classification
 ---
-
-# Thyroid Nodule Classification – EfficientNetV2-S (AUC-Optimized v2)
-
-This model was trained on the TN5000 thyroid ultrasound dataset for binary classification of thyroid nodules (Benign vs Malignant), **optimized for AUC-ROC**.
-
-## Key Training Decisions
-
-- **No class weighting in BCE loss**: AUC is a ranking metric; class weighting distorts probability calibration and hurts discriminative power.
-- **WeightedRandomSampler**: Used to ensure balanced batches during training.
-- **EMA decay 0.999**: For stable validation/test predictions.
-
-## Training Configuration
-
-| Parameter | Value |
-|-----------|-------|
-| Backbone | tf_efficientnetv2_s.in1k (timm, ImageNet pretrained) |
-| Input size | 384×384 |
-| Batch size | 64 |
-| Epochs | 11 (early stopped) |
-| LR head | 0.0001 |
-| LR backbone | 1e-05 |
-| Weight decay | 0.0001 |
-| Warmup | 3 epochs |
-| Scheduler | Cosine annealing |
-| EMA decay | 0.999 |
-| Loss | BCE (no pos_weight) |
-| Optimization metric | AUC-ROC |
+# Thyroid Nodule Classification - EfficientNetV2-S (AUC-Optimized v4)
+Frozen backbone, deeper head with Dropout 0.5. Optimized for AUC-ROC.
 
 ## Test Set Performance
-
 | Metric | Value | 95% CI |
 |--------|-------|--------|
-| Sensitivity | 0.9111 | [0.8881, 0.9307] |
-| Specificity | 0.1115 | [0.0765, 0.1554] |
-| PPV | 0.7359 | [0.7059, 0.7644] |
-| NPV | 0.3158 | [0.2242, 0.4192] |
-| AUC-ROC | 0.5195 | [0.4772, 0.5549] |
-
+| Sensitivity | 0.1751 | [0.1482, 0.2046] |
+| Specificity | 0.9294 | [0.8919, 0.9569] |
+| PPV | 0.8707 | [0.8055, 0.9204] |
+| NPV | 0.2931 | [0.2627, 0.3249] |
+| AUC-ROC | 0.6835 | [0.6467, 0.7199] |
 ## Citation
-
-Yu, Xiaoxian et al. "TN5000: An Ultrasound Image Dataset for Thyroid Nodule Detection and Classification." Scientific Data (Nature), 2025.
-
-<!-- ml-intern-provenance -->
-## Generated by ML Intern
-
-This model repository was generated by [ML Intern](https://github.com/huggingface/ml-intern), an agent for machine learning research and development on the Hugging Face Hub.
-
-- Try ML Intern: https://smolagents-ml-intern.hf.space
-- Source code: https://github.com/huggingface/ml-intern
-
-## Usage
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-model_id = 'Johnyquest7/Thyroid_EfficientNetV2'
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id)
-```
-
-For non-causal architectures, replace `AutoModelForCausalLM` with the appropriate `AutoModel` class.
+Yu, Xiaoxian et al. "TN5000..." Scientific Data (Nature), 2025.

best_model.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:15482835c21c210cee07fa1abcfa57738e34d45140074311f91b883e55fd7c98
-size 324288274
+oid sha256:1475972504c0a9df090290d68e2fea6ee1d719a1de25e0e2055ae0683d3fbf58
+size 82917646

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ff41a1f37ee9f32c2d78c0371fd7cc230ec8187344f4d72c8c7c692d9ce5d65
-size 81611074
+oid sha256:50344f12c08700a55344eb5c64cd45994bf32e878b45715ada7f0c4711344d95
+size 82919434

results.json

@@ -1,48 +1,46 @@
 {
-  "best_epoch": 1,
-  "best_val_auc": 0.4818346666666667,
+  "best_epoch": 24,
+  "best_val_auc": 0.6955626666666666,
   "test_metrics": {
-    "sensitivity": 0.9110807113543091,
+    "sensitivity": 0.17510259917920656,
     "sensitivity_ci": [
-      0.8880679545923451,
-      0.930704677143796
+      0.14822370178949582,
+      0.2046325480923097
     ],
-    "specificity": 0.11152416356877323,
+    "specificity": 0.929368029739777,
     "specificity_ci": [
-      0.07652180042423258,
-      0.155368432564055
+      0.8918951222927511,
+      0.9569420548075194
     ],
-    "ppv": 0.7359116022099448,
+    "ppv": 0.8707482993197279,
     "ppv_ci": [
-      0.7058993124376329,
-      0.7643793122110114
+      0.8055358886428969,
+      0.9203548937021385
     ],
-    "npv": 0.3157894736842105,
+    "npv": 0.29308323563892147,
     "npv_ci": [
-      0.22421979057887004,
-      0.41920736370136785
+      0.26271020298013675,
+      0.3248933746402657
     ],
-    "auc": 0.5194773163004287,
+    "auc": 0.6834783537345084,
     "auc_ci": [
-      0.4771984664530449,
-      0.5548698473128044
+      0.6466884099392298,
+      0.7199268327692855
     ],
-    "tp": 666,
-    "tn": 30,
-    "fp": 239,
-    "fn": 65,
+    "tp": 128,
+    "tn": 250,
+    "fp": 19,
+    "fn": 603,
     "threshold": 0.5
   },
   "config": {
     "backbone": "tf_efficientnetv2_s.in1k",
+    "frozen": true,
     "img_size": 384,
     "batch_size": 64,
-    "epochs_trained": 11,
-    "lr_head": 0.0001,
-    "lr_backbone": 1e-05,
-    "weight_decay": 0.0001,
-    "ema_decay": 0.999,
-    "optimization_metric": "auc",
-    "pos_weight": 1.0
+    "epochs_trained": 39,
+    "lr_head": 0.0005,
+    "weight_decay": 0.001,
+    "optimization_metric": "auc"
   }
 }