segresnet

This project compares the performance of UNet, ResUNet, SegResNet, and UNETR architectures on the 2017 LiTS dataset for liver tumor segmentation. We evaluate segmentation accuracy using the DICE score to identify key factors for effective tumor segmentation.

🧠 3D Brain Tumor Segmentation with MONAI | BRATS 2020 Baseline UNet, UNet++, SegResNet 🚀

3D multi-label brain tumor segmentation from multimodal MRI (FLAIR, T1w, T1gd, T2w) using Decathlon data. End-to-end workflow with 3D SegResNet, Dice loss, and Mean Dice evaluation for glioma sub-region segmentation (edema, core, necrotic/cystic, enhancing). Focused on complex, heterogeneous tumor localization in clinical neuroimaging.

Brain tumor segmentation on BraTS-2023 dataset using nnUNet, SegResNet, and SwinUNETR. Models trained and evaluated on MRI scans to segment tumor sub-regions: WT, TC, ET, and RC, with fold-wise performance visualizations.

Comparative study of 3D deep learning architectures (U-Net, Attention U-Net, SegResNet, Separate-Encoder U-Net) for ischemic stroke lesion segmentation on ISLES 2022

3D glioma MRI segmentation with Centralized, FedAvg and FedProx training, plus encoder-based patient retrieval on BraTS.

Longitudinal brain-MRI disease-evolution platform. Tracks how tumour findings change across serial studies: 3D SegResNet segmentation (0.92 Dice whole tumour, held-out BraTS), registration, cross-time lesion matching, change detection with propagated uncertainty, evidence-linked 3D timeline. Every answer traces to a measurement. Research prototype.