Valentino, Marika (2024) Computational and Quantitative Phase Microscopy for label-free Cells and Tissues Analysis. [Tesi di dottorato]

Documento PDF PhD_Thesis_Marika_Valentino_XXXVII.pdf Visibile a [TBR] Amministratori dell'archivio Download (373MB) | Richiedi una copia

Abstract

Quantitative Phase Microscopy (QPM) is emerging as a novel way of observing biological cells and tissues. Differently from more conventional paradigms that rely on the fluorescence emission readout, in QPM sample staining/labelling is not required. Image contrast is endogenous and is provided by the optical phase delay occurring when light crosses biological samples. This concept overturns conventional microscopy and in particular fluorescence microscopy methods. Neither fluorescent markers nor external dyes are needed. Only light plays a crucial role. Hence, the staining-induced ambiguities, e.g. phototoxicity, photobleaching, and lab/protocol-dependence can be avoided allowing more accurate and objective diagnoses. Among QPM techniques, Digital Holographic Microscopy (DHM) and Fourier Ptychographic Microscopy (FPM) gathers quantitative, real-time and label-free images with advanced optical performance. Moreover, the recent advancements in computational microscopy unlocked a fruitful integration between QPM methodologies and artificial intelligence (AI). This novel interdisciplinary framework can enhance diagnostic capabilities, facilitating rapid analysis of large datasets and supporting timely, accurate decision-making in clinical practice. This PhD Thesis aims to develop and apply novel computational and AI-aided QPM methods to foster advanced cells and tissue analysis, further contributing to the ongoing process that is positioning them as a valuable diagnostic method used in clinical practice. More than encouraging results are obtained in this thesis, where cells and tissues are characterized in label-free, quantitative and multi-scale manner. The assessment of the full computational pipeline, including phase image reconstruction algorithms, segmentation algorithms, classification and image transfer algorithms, enables novel modes for single cell analysis and tissue characterization by using DHM and FPM. AI-driven methods, such as real-time emulation of the FPM phase retrieval, histopathology tissue classification and virtual staining, are implemented and discussed, demonstrating the powerful synergy between QPM and AI in modern diagnostics. These findings suggest that DHM and FPM will play a more central role in revolutionizing personalized medicine, through accurate single-cell analysis and diagnostics.

Downloads

Actions (login required)

Modifica documento