Upload folder using huggingface_hub

README.md

@@ -1,45 +1,33 @@
 ---
 license: cc-by-4.0
 tags:
-- medical-imaging
-- ultrasound
-- thyroid
-- classification
-- efficientnet
-- ml-intern
+  - medical-imaging
+  - ultrasound
+  - thyroid
+  - classification
+  - resnet
 datasets:
-- Johnyquest7/TN5000-thyroid-nodule-classification
+  - Johnyquest7/TN5000-thyroid-nodule-classification
 ---
-# Thyroid Nodule Classification - EfficientNetV2-S (AUC-Optimized v4)
-Frozen backbone, deeper head with Dropout 0.5. Optimized for AUC-ROC.
+# Thyroid Nodule Classification - ResNet-18 (PEMV-Style Correct)
+Trained on TN5000 with exact PEMV paper recipe, optimized for AUC-ROC.
+
+## Key Recipe Differences from Failed Runs
+- No ImageNet normalization (only ToTensor to [0,1])
+- CrossEntropyLoss with 2 logits (not BCE with 1 logit)
+- ResNet-18 (proven 85.68% accuracy baseline on TN5000)
+- AdamW lr=1e-4, wd=0.05, batch=16, 128x128
+- Constant LR for 200 epochs (no scheduler)
 
 ## Test Set Performance
 | Metric | Value | 95% CI |
 |--------|-------|--------|
-| 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..." 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.
+| Accuracy | 0.8891 | - |
+| Sensitivity | 0.9175 | [0.8948, 0.9366] |
+| Specificity | 0.8182 | [0.7685, 0.8611] |
+| PPV | 0.9266 | [0.9048, 0.9447] |
+| NPV | 0.7986 | [0.7481, 0.8430] |
+| AUC-ROC | 0.9313 | [0.9125, 0.9483] |
+
+## References
+- PEMV-Thyroid (arXiv:2603.28315): Prototype-Enhanced Multi-View Learning for Thyroid Nodule Ultrasound Classification

best_model.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1475972504c0a9df090290d68e2fea6ee1d719a1de25e0e2055ae0683d3fbf58
-size 82917646
+oid sha256:48543f2595e272d6d1cbcd9f28ec9877395bc72009585e4dc0ba29cee8c906a0
+size 44790904

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:50344f12c08700a55344eb5c64cd45994bf32e878b45715ada7f0c4711344d95
-size 82919434
+oid sha256:b83f55db67889c856d363362ad29442738742adf3a9b2e1c791b997b5cc47d27
+size 44790706

results.json

@@ -1,46 +1,46 @@
 {
-  "best_epoch": 24,
-  "best_val_auc": 0.6955626666666666,
+  "best_epoch": 12,
+  "best_val_auc": 0.9318691478479817,
   "test_metrics": {
-    "sensitivity": 0.17510259917920656,
+    "accuracy": 0.8891108891108891,
+    "sensitivity": 0.9174825174825175,
     "sensitivity_ci": [
-      0.14822370178949582,
-      0.2046325480923097
+      0.8948475523330561,
+      0.9365925537099365
     ],
-    "specificity": 0.929368029739777,
+    "specificity": 0.8181818181818182,
     "specificity_ci": [
-      0.8918951222927511,
-      0.9569420548075194
+      0.7685054712783398,
+      0.8611264590466025
     ],
-    "ppv": 0.8707482993197279,
+    "ppv": 0.9265536723163842,
     "ppv_ci": [
-      0.8055358886428969,
-      0.9203548937021385
+      0.9047977037903637,
+      0.9446634522191668
     ],
-    "npv": 0.29308323563892147,
+    "npv": 0.7986348122866894,
     "npv_ci": [
-      0.26271020298013675,
-      0.3248933746402657
+      0.7480838301457988,
+      0.8430321946896866
     ],
-    "auc": 0.6834783537345084,
+    "auc": 0.9312704777739742,
     "auc_ci": [
-      0.6466884099392298,
-      0.7199268327692855
+      0.9124902286540131,
+      0.9482509242450335
     ],
-    "tp": 128,
-    "tn": 250,
-    "fp": 19,
-    "fn": 603,
-    "threshold": 0.5
+    "tp": 656,
+    "tn": 234,
+    "fp": 52,
+    "fn": 59
   },
   "config": {
-    "backbone": "tf_efficientnetv2_s.in1k",
-    "frozen": true,
-    "img_size": 384,
-    "batch_size": 64,
-    "epochs_trained": 39,
-    "lr_head": 0.0005,
-    "weight_decay": 0.001,
+    "backbone": "resnet18.a1_in1k",
+    "img_size": 128,
+    "batch_size": 16,
+    "epochs_trained": 32,
+    "lr": 0.0001,
+    "weight_decay": 0.05,
+    "normalization": "none",
     "optimization_metric": "auc"
   }
 }