Mélanie Roschewitz, PhD

About me

I’m an ML researcher interested in the reliability, safety and fairness of AI systems for healthcare.

Background

I started my studies with a B.Sc. in Mathematics as well as a postgraduate diploma in Statistics from the University of Strasbourg (France). Later, I graduated from ETH Zurich with a M.Sc. in Data Science, focusing on medical applications of machine learning. I then joined Microsoft Research Cambridge (UK), where I was an Applied Researcher in their medical imaging team, focusing on developing ML models for personalized treatment as well as on open-source efforts in ML for healthcare. In October 2021, I then started my PhD at at Imperial College London, advised by Prof. Ben Glocker. During my PhD, I was a 2024 recipient of the Google PhD fellowship in Health & Bioscience and of the Imperial College President’s scholarship. And I also joined the Kheiron Medical Technologies ML research team for an internship from April to October 2022.

News

Jul 01, 2025	I’ve successfully graduated from my PhD, with my thesis: ‘Towards robust and reliable disease classification in medical imaging’ 🎓🎉
Jun 22, 2025	Our papers Automatic dataset shift identification to support safe deployment of medical imaging AI and CF-Seg: Counterfactuals meet Segmentation have been accepted at MICCAI 2025!
Jun 16, 2025	Our paper Robust image representations with counterfactual contrastive learning has been published in Medical Image Analysis.
Nov 18, 2024	I’ve been awarded a Google PhD Fellowship in Health & Bioscience 🎉
Nov 12, 2024	Have been recognised as top reviewer for NeurIPS 2024

Selected publications

MICCAI

Automatic dataset shift identification to support safe deployment of medical imaging AI

Mélanie Roschewitz , Raghav Mehta , Charles Jones , and 1 more author

In International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), (in press) , Sep 2025

HTML Code
MedIA

Robust image representations with counterfactual contrastive learning

Mélanie Roschewitz , Fabio De Sousa Ribeiro , Tian Xia , and 2 more authors

Medical Image Analysis, Sep 2025

HTML

NatureComms

Automatic correction of performance drift under acquisition shift in medical image classification

Mélanie Roschewitz , Galvin Khara , Joe Yearsley , and 7 more authors

Nature Communications, Oct 2023

Bib HTML Code

@article{roschewitz2023automatic,
  title = {Automatic correction of performance drift under acquisition shift in medical image classification},
  author = {Roschewitz, M{\'e}lanie and Khara, Galvin and Yearsley, Joe and Sharma, Nisha and James, Jonathan J and Ambr{\'o}zay, {\'E}va and Heroux, Adam and Kecskemethy, Peter and Rijken, Tobias and Glocker, Ben},
  journal = {Nature Communications},
  volume = {14},
  number = {1},
  pages = {6608},
  year = {2023},
  month = oct,
  publisher = {Nature Publishing Group UK London},
  selected = true,
}

NatureComms

Active label cleaning for improved dataset quality under resource constraints

Mélanie Bernhardt , Daniel C Castro , Ryutaro Tanno , and 8 more authors

Nature communications, Mar 2022

Bib HTML Code

@article{bernhardt2022active,
  title = {Active label cleaning for improved dataset quality under resource constraints},
  author = {Bernhardt, M{\'e}lanie and Castro, Daniel C and Tanno, Ryutaro and Schwaighofer, Anton and Tezcan, Kerem C and Monteiro, Miguel and Bannur, Shruthi and Lungren, Matthew P and Nori, Aditya and Glocker, Ben and others},
  journal = {Nature communications},
  volume = {13},
  number = {1},
  pages = {1161},
  year = {2022},
  month = mar,
  publisher = {Nature Publishing Group UK London},
  selected = true,
}

NatureMed

Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithms

Mélanie Bernhardt , Charles Jones , and Ben Glocker

Nature Medicine, Jun 2022

Bib HTML

@article{bernhardt2022potential,
  title = {Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithms},
  author = {Bernhardt, M{\'e}lanie and Jones, Charles and Glocker, Ben},
  journal = {Nature Medicine},
  volume = {28},
  number = {6},
  pages = {1157--1158},
  month = jun,
  year = {2022},
  publisher = {Nature Publishing Group US New York},
  selected = true
}

TMLR

Failure Detection in Medical Image Classification: A Reality Check and Benchmarking Testbed

Mélanie Bernhardt , Fabio De Sousa Ribeiro , and Ben Glocker

Transactions on Machine Learning Research, Oct 2022

Bib HTML Code

@article{bernhardt2022failure,
  title = {Failure Detection in Medical Image Classification: A Reality Check and Benchmarking Testbed},
  author = {Bernhardt, M{\'e}lanie and Ribeiro, Fabio De Sousa and Glocker, Ben},
  journal = {Transactions on Machine Learning Research},
  year = {2022},
  selected = true,
  month = oct
}