Extrachromosomal DNA leaves a reproducible histomorphologic footprint detectable in routine pathology

This repository contains the implementation of AMIE, an end-to-end multiple instance learning (MIL) framework to detect extrachromosomal DNA (ecDNA) from routine haematoxylin and eosin (H&E) whole-slide images (WSIs).

📢 Updates

• 02/03/2026: Our manuscript “Detecting Extrachromosomal DNA from Routine Histopathology” is now available on BioRxiv.

Usage Guide

Prerequisites

• PyTorch 2.10
• OpenSlide 1.4
• OpenCV 4.13
• NumPy/pandas/SciPy/scikit-learn

Installation

1. Clone the repository:

git clone https://github.com/manwaarkhd/amie.git
cd amie

2. Set up a Python virtual environment using venv (optional but recommended):

python -m venv env
source env/bin/activate # on windows use: env\Scripts\activate

3. Install the required packages:

pip install -r requirements.txt

Acknowledgements

The authors acknowledge the support of the Ministry of Science and Culture of Lower Saxony through funds from the program zukunft.niedersachsen of the Volkswagen Foundation for the CAIMed – Lower Saxony Center for Artificial Intelligence and Causal Methods in Medicine project (grant no. ZN4257). The authors acknowledge Hannover Medical School for providing MHH-HPC resources and technical support that have contributed to the research results reported within this paper.

Cite

If you find our work useful in your research, please consider citing our paper:

@article {Khalid2026.02.27.708546,
	author = {Khalid, Muhammad Anwaar and Gratius, Michael and Brown, Christopher and Younis, Raneen and Ahmadi, Zahra and Chavez, Lukas},
	title = {Detecting Extrachromosomal DNA from Routine Histopathology},
	year = {2026},
	doi = {10.64898/2026.02.27.708546},
	publisher = {Cold Spring Harbor Laboratory},
	journal = {bioRxiv}
}