LightningMedSeg3D: Trained Weights of Nine 3D Medical Image Segmentation Networks on BTCV and MSD Task03 (Liver)

Trained model weights for nine 3D medical image segmentation architectures (UNet, VNet, ResUNet, Attention U-Net, UNet++, UNETR, SwinUNETR, MedFormer and SegFormer), produced with the LightningMedSeg3D framework (https://github.com/Removirt/LightningMedSeg3D) under identical training protocols. Two datasets are covered: the Beyond the Cranial Vault (BTCV) abdominal multi-organ CT benchmark, and Task03 (Liver) of the Medical Segmentation Decathlon (liver and liver-tumour segmentation). Each architecture is provided as a PyTorch state_dict (.pth) for both datasets (18 files in total). Per-architecture quantitative results (Dice, Normalized Surface Dice, MASD, Hausdorff Distance and Relative Volume Difference) and publication-quality comparison figures are included. Funding: Spanish Ministerio de Ciencia e Innovación and Agencia Estatal de Investigación, grants PID2022-142709OB-C21 and PID2022-142709OA-C22.