Netherlands
Deep learning methods to predict cancer therapy response from histopathology slides. I explain my PhD research in a three minute pitch that won the University of Amsterdam wide pitch competition here: https://www.youtube.com/watch?v=_Se6Ag3e20E See my work: - DeepSMILE: Self-supervised heterogeneity-aware multiple instance learning for DNA damage response defect classification directly from H&E whole-slide images (https://arxiv.org/abs/2107.09405) - WeakSTIL: Weak whole-slide image level stromal tumor infiltrating lymphocyte scores are all you need (https://arxiv.org/abs/2109.05892) - ECTIL: Label-efficient Computational Tumour Infiltrating Lymphocyte (TIL) assessment in breast cancer: Multicentre validation in 2,340 patients with breast cancer (https://arxiv.org/abs/2501.14379, https://github.com/nki-ai/ectil) - From Explainable to Explained AI: Ideas for Falsifying and Quantifying Explanations (https://arxiv.org/abs/2508.09205, https://github.com/NKI-AI/x2x) - HRD-related morphology discovery in breast cancer by controlling for confounding factors (https://www.sciencedirect.com/science/article/pii/S2666379122004372) - Generalizable biomarker prediction from cancer pathology slides with self-supervised deep learning: A retrospective multi-centric study (https://www.cell.com/cell-reports-medicine/pdf/S2666-3791(23)00086-1.pdf) Appearances: - Conference speaker at ESMO MAP Amsterdam 2022 - Speaker at the The Intelligent Health Summit: Maximizing the clinical value of whole-slide images of cancer tissue using computer vision: From clinical problems to novel deep learning solutions - Oral at SPIE Medical Imaging 2022, presenting WeakSTIL (proceedings and presentation to be published) - Invited speaker at PathAI's "Let's talk AI" speaker series. Discussed DeepSMILE, WeakSTIL, SSL in pathology. MSc thesis supervision for Data Science, Bioinformatics, and AI Reviewer for MICCAI '21 '25, MEDIA, CVPR '22 '23, Cell Reports Medicine, Cancer Cell, ECCV '22, TMLR, NPJ
Research assistant for development of chatbot to test the psychological influence of prosody in speech: Examining the persuasiveness of text and voice agents: prosody aligned with information structure increases human-likeness, perceived personalisation and brand attitude (https://www.tandfonline.com/doi/full/10.1080/0144929X.2024.2420871) Chatbot in React using Botonic hosted on AWS amplify, extended the CART python backend, hosted on AWS Elastic Beanstalk PhD student @ UvA
Deep Learning course
Visiting Jakob Kather’s lab to research improved AI methods for clinically actionable target prediction from H&E whole-slide images of cancer tissue