| --- |
| language: en |
| license: mit |
| library_name: pytorch |
| tags: |
| - medical-imaging |
| - mri |
| - 3d-resnet |
| - monai |
| - oncology |
| - breast-cancer |
| datasets: |
| - odelia |
| metrics: |
| - roc_auc |
| pipeline_tag: image-classification |
| --- |
| |
| # 3D ResNet-18 for ODELIA MRI Classification |
|
|
| This repository contains the weights for a 3D ResNet-18 model trained for the ODELIA dataset (multi-parametric MRI). The model was developed to classify medical images into 3 distinct classes using 5-fold cross-validation. |
|
|
| ## Model Description |
|
|
| - **Developed by:** THOUAN Simon |
| - **Model type:** 3D Convolutional Neural Network (ResNet-18 architecture) |
| - **Framework:** PyTorch & MONAI |
| - **Task:** Multi-class 3D Image Classification (3 classes) |
| - **Input:** 5 MRI sequences (Pre, Post_1, Post_2, Sub_1, T2) concatenated as channels. |
| - **Input Size:** (128, 128, 64) |
| |
| ## Architecture Details |
| |
| The model is a 3D adaptation of the ResNet architecture provided by the **MONAI** library: |
| - **Blocks:** Basic block |
| - **Layers:** [2, 2, 2, 2] (Equivalent to ResNet-18) |
| - **In-planes:** [64, 128, 256, 512] |
| - **Input Channels:** 5 |
| - **Spatial Dimensions:** 3D |
| |
| ## Training Procedure |
| |
| The model was trained on a high-performance computing cluster (IDUN) using the following configuration: |
| |
| ### Hyperparameters |
| | Parameter | Value | |
| | :--- | :--- | |
| | **Optimizer** | Adam | |
| | **Learning Rate** | 1e-4 | |
| | **Loss Function** | CrossEntropyLoss | |
| | **Batch Size** | 4 (Training) / 2 (Validation) | |
| | **Epochs** | 50 | |
| | **Validation** | Every 2 epochs | |
| |
| ### Cross-Validation Strategy |
| The dataset was split into 5 folds (A, B, C, D, E) to ensure robustness. |
| - **Fold 0:** Val = A, Train = B+C+D+E |
| - **Fold 1:** Val = B, Train = A+C+D+E |
| - *(Continuing for all 5 folds)* |
| |
| ### Preprocessing (MONAI Transforms) |
| - **Resizing:** All volumes resized to $128 \times 128 \times 64$. |
| - **Normalization:** Intensity scaling for each sequence. |
| - **Concatenation:** The 5 MRI sequences are stacked into a single 5-channel tensor. |
| |
| ## Evaluation Results |
| |
| The primary metric used is **ROC AUC** (Area Under the Receiver Operating Analytic Curve). |
| Detailed performance graphs and confusion matrices for the ensemble can be found in the associated [GitHub Repository](https://github.com/THOUAN-Simon/ODELIA_Challenge_CV_DL)). |
| |
| ## How to Load the Model |
| |
| ```python |
| import torch |
| from monai.networks.nets import ResNet |
| |
| # Initialize architecture |
| model = ResNet( |
| block="basic", |
| layers=[2, 2, 2, 2], |
| block_inplanes=[64, 128, 256, 512], |
| n_input_channels=5, |
| num_classes=3, |
| spatial_dims=3 |
| ) |
| |
| # Load weights |
| state_dict = torch.load("weights/best_resnet_odelia_fold0.pth", map_location="cpu") |
| model.load_state_dict(state_dict) |
| model.eval() |
|
|
| ``` |
| |
| ## Limitations & Ethical Considerations |
| This model is for research purposes only. It was trained on the ODELIA proprietary dataset. Predictions should not be used for clinical diagnosis without professional medical supervision. |
| |
