Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,61 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
model_name: brats-segresnet-monai
|
| 3 |
+
tags:
|
| 4 |
+
- image-segmentation
|
| 5 |
+
- segresnet
|
| 6 |
+
- medical-imaging
|
| 7 |
+
- brats
|
| 8 |
+
- fp16
|
| 9 |
+
license: apache-2.0
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
+
pipeline_tag: image-segmentation
|
| 13 |
+
library_name: monai
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# 🧠 SegResNet — BraTS Brain Tumor Segmentation (3D, FP16)
|
| 17 |
+
|
| 18 |
+
This model fine-tunes **SegResNet** for **3D brain tumor segmentation** (Tumor Core, Whole Tumor, Enhancing Tumor) using the
|
| 19 |
+
[BraTS2020 Dataset](https://www.kaggle.com/datasets/awsaf49/brats20-dataset-training-validation).
|
| 20 |
+
The model was trained with **mixed precision (FP16)** using MONAI and Accelerate for efficient multi-GPU training.
|
| 21 |
+
|
| 22 |
+
---
|
| 23 |
+
|
| 24 |
+
## ⚙️ Configuration
|
| 25 |
+
|
| 26 |
+
| Attribute | Value |
|
| 27 |
+
|------------|-------|
|
| 28 |
+
| **Base Model** | `SegResNet` (MONAI) |
|
| 29 |
+
| **Dataset** | [BraTS2020 (Kaggle)](https://www.kaggle.com/datasets/awsaf49/brats20-dataset-training-validation) |
|
| 30 |
+
| **Epochs** | 150 |
|
| 31 |
+
| **Batch Size (per GPU)** | 2 |
|
| 32 |
+
| **ROI Size** | 128×128×128 |
|
| 33 |
+
| **Optimizer** | Adam (LR=0.0001, WD=1e-05) |
|
| 34 |
+
| **Precision** | FP16 (mixed=True) |
|
| 35 |
+
| **Framework** | MONAI + Accelerate |
|
| 36 |
+
| **Device** | Multi-GPU (num_processes=2) |
|
| 37 |
+
|
| 38 |
+
---
|
| 39 |
+
|
| 40 |
+
## 🩺 Example Segmentation
|
| 41 |
+
|
| 42 |
+
### 2D Prediction
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
### 3D Prediction
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
---
|
| 49 |
+
|
| 50 |
+
## 📈 Results (Final Test Metrics)
|
| 51 |
+
|
| 52 |
+
| Metric | Average | Tumor Core (TC) | Whole Tumor (WT) | Enhancing Tumor (ET) |
|
| 53 |
+
|--------|---------|-----------------|------------------|----------------------|
|
| 54 |
+
| **Dice** | 0.789 | 0.775 | 0.874 | 0.719 |
|
| 55 |
+
| **Precision** | 0.847 | 0.856 | 0.938 | 0.749 |
|
| 56 |
+
| **Recall** | 0.771 | 0.762 | 0.839 | 0.713 |
|
| 57 |
+
| **Accuracy** | 0.987 | 0.999 | 0.999 | 0.964 |
|
| 58 |
+
| **Loss** | 0.192 | - | - | - |
|
| 59 |
+
|
| 60 |
+
> FP16 mixed precision enabled faster training while maintaining high segmentation accuracy.
|
| 61 |
+
> No data augmentation was used to preserve original image fidelity.
|