Title: Semantic Unmixing of Fluorescence Microscopy Data using Deep Learning

Abstract: Fluorescence microscopy is limited by optics, fluorophore chemistry, and photon exposure, forcing trade-offs in speed, resolution, and depth. In this talk, I will discuss my PhD research that addresses these challenges. Several deep-learning-based computational multiplexing techniques will be discussed which enhances imaging of multiple cellular structures within a single fluorescent channel, enabling faster imaging with less photon exposure. More concretely, I focus on unmixing superimposed structures—such as Nucleus and Tubulin—into separate images. One approach employs a Hierarchical Variational Autoencoder (H-VAE) inspired architecture, allowing sampling of diverse, plausible predictions from a trained posterior. Additionally, I present methods for uncertainty quantification and calibration to improve reliability. To handle cases where one structure dominates in the superimposed image, we leverage the inductive bias of flow-matching techniques and develop an image-unmixing framework using InDI. These advances aim to push the boundaries of fluorescence microscopy through computational innovation.

About the Speaker: Ashesh is a final year PhD student in the Computer Science department at TU Dresden, Germany. He has done all his PhD research at Florian Jug’s lab, Computational Biology Center, Human Technopole (HT), Milan, Italy. His PhD project is an image decomposition task of splitting a superimposed (fluorescence microscopy) image into its constituent channels. With a couple of publications in top Comuter vision conferences and with a few pre-prints, his PhD work has provided a credible solution. Due to these works, he got a 8700€ EMBO (European Molecular Biology Organization) grant to visit ENS de Lyon, France to work on self-supervised finetuning and uncertainty quantification for his PhD problem. He also won the best oral presentation award for my PhD work at HT PhD-&-Postdoc Symposium 2024. Previously, he received a Dual (B.Tech+M.Tech) degree in Computer Science in 2015, from IIT Delhi, India. He has more than 3 years of experience working in Industry, mostly as a Data Scientist. He also worked at National Taiwan University, Taipei, Taiwan as a Research Assistant for about an year under Prof. Hsuan-Tien Lin. During his time there, he initiated research in the lab on multiple Computer Vision problems in the lab, culminating in published works.