ChronoSurv: A Clinical Pathway-Guided Graph Framework for Multimodal Survival Analysis 🧪🔬🎯

MICCAI 2026

---

🧩 Method Overview

We propose ChronoSurv, a Graph Neural Network architecture for multimodal survival prediction that models the clinical pathway as a directed heterogeneous graph with temporal progression. In the codebase, the model is exposed as chrono_surv with the matching datamodule type UnifiedHNC_ChronoSurv.

---

🚀 Getting Started

Installation

Requirements: Python 3.12+

# Clone the repository
git clone https://github.com/MICS-Lab/ChronoSurv.git
cd ChronoSurv

# Install dependencies
pip install -r requirements.txt

📂 Dataset Preparation

This project uses two head and neck cancer cohorts:

1. HANCOCK (primary dataset)

Download the HANCOCK dataset and place it in ./data/HANCOCK/.

data/HANCOCK/
├── StructuredData/
│   ├── clinical_data.json
│   ├── blood_data.json
│   └── pathological_data.json
├── TextData/
│   ├── histories_english/
│   ├── surgery_descriptions_english/
│   └── reports_english/
├── TMA_CellDensityMeasurements/
│   └── TMA_celldensity_measurements.csv
├── WSI_LymphNode/
│   └── h5_files/
├── WSI_PrimaryTumor/
│   └── WSI_PrimaryTumor_*/
└── Split/
    └── folds_5.csv

2. TCGA-HNSC (secondary dataset)

Download TCGA-HNSC clinical and WSI data and place it in ./data/TCGA-HNSC/.

data/TCGA-HNSC/
├── clinical_data.json
├── WSI_PrimaryTumor/
│   └── h5_files/
└── Split/
    └── folds_5.csv

Build K-Folds (Optional)

# HANCOCK
python main.py folds --dataset hancock --data_root ./data/HANCOCK --n_folds 5 --random_seed 42

# TCGA-HNSC
python main.py folds --dataset tcga --data_root ./data/TCGA-HNSC --n_folds 5 --random_seed 42

---

Training

Basic Usage

python main.py train --config config/chrono_surv.yaml

# For all available options:
python main.py train --help

---

Evaluation

python main.py eval --checkpoint-dirs [checkpoint-dirs]

---

🙌 Acknowledgments

We acknowledge Kist et al. 2024 for making the HANCOCK dataset available.

Useful Links

• HANCOCK Challenge
• TCGA-HNSC
• BioClinicalBERT

---

🔗 Citation

Important

This project is based on the work by Miccinilli and Di Piazza 2026. If you use this code in your research, we would appreciate reference to the following paper:

@inproceedings{md_chronosurv_2026,
  author    = {Hugo Miccinilli and Theo Di Piazza},
  title     = {ChronoSurv: A Clinical Pathway-Guided Graph Framework for Multimodal Survival Analysis},
  booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
  year      = {2026},
}