Multimodal Prostate Cancer BCR Prediction

This repository provides the pretrained models for biochemical recurrence (BCR) prediction in prostate cancer using:

Multiparametric MRI (mpMRI) features
Clinical variables
A multimodal survival model trained with a ranking-based loss (C-index–oriented)

The models are released as PyTorch checkpoints and hosted on Hugging Face for easy reuse.

from huggingface_hub import hf_hub_download
import torch
from src.models import SurvivalModelMM

# Download model
model_path = hf_hub_download(repo_id="farzadbz/BCR_prediction_model", filename="your desired model path")
ckpt = torch.load(model_path, map_location="cpu")
model = SurvivalModelMM(modalities=modalities, d_emb=embed_dim).to(device)
model.load_state_dict(ckpt)