BEGIN:VCALENDAR VERSION:2.0 PRODID:Icfo X-PUBLISHED-TTL:P1W BEGIN:VEVENT UID:69d4b69aa54fd DTSTART:20251121T090000Z SEQUENCE:0 TRANSP:OPAQUE LOCATION:ICFO Auditorium SUMMARY:ICFO | LUIS FELIPE MORALES CURIEL CLASS:PUBLIC DESCRIPTION:Bioluminescence microscopy presents a powerful alternative to f luorescence imaging by eliminating the need for external illumination\, th ereby avoiding issues such as phototoxicity\, photobleaching\, and backgro und autofluorescence. However\, the inherently low photon output of lucife rase-based reporters significantly restricts the signal-to-noise ratio (SN R)\, as well as the achievable spatial and temporal resolution&mdash\;chal lenges that are especially pronounced in dynamic or volumetric biological imaging. This thesis addresses these limitations by introducing a deep lea rning-driven imaging pipeline designed to enhance bioluminescence microsco py at both the data acquisition and image reconstruction stages. \;\nO ur strategy integrates optical system design with advanced neural networks to enable rapid\, high-resolution 3D imaging under extremely low-light co nditions. We engineered a custom microscope featuring a highly compact opt ical axis and paired it with a single-photon sensitive camera\, significan tly boosting the SNR of bioluminescent images. To achieve fast volumetric imaging\, we incorporated light field microscopy (LFM) and Fourier light f ield microscopy (FLFM)\, enabling single-shot 3D acquisition while improvi ng axial and lateral resolution via Fourier-domain filtering. The primary objective of this work is to demonstrate how deep learning can substantial ly enhance bioluminescence microscopy\, pushing the technique beyond its t raditional limits in both 2D and 3D imaging.\nAt the core of our approach is a suite of convolutional neural networks specifically trained on biolum inescent data. Using both synthetic and experimental datasets\, we designe d and trained models capable of extracting meaningful information from low -SNR raw data\, recovering otherwise lost details and offering deeper insi ght into the biological sample. The models developed in this thesis cover key tasks such as denoising and reconstruction of wide-field\, light field \, and Fourier light field bioluminescent images. Together\, they form a m odular\, learnable pipeline that significantly elevates the performance of bioluminescence microscopy in terms of both quality and speed.\nWe valida te our system using live biological samples\, including Caenorhabditis ele gans\, mouse stem cells\, and zebrafish embryos\, capturing neuronal activ ity and intracellular dynamics at subsecond timescales. By placing deep le arning at the heart of the imaging process\, this work establishes a new p aradigm for bioluminescence microscopy\, transforming a traditionally low- SNR modality into a robust tool for fast\, high-resolution\, and label-spe cific imaging in living organisms.\nFriday\, November 21\, 10:00 h. ICFO A uditorium \nThesis Director: Prof. Dr. Michael Krieg DTSTAMP:20260407T074738Z END:VEVENT END:VCALENDAR