Optical microscopy is a powerful tool employed in the field of biology to investigate the microscopic properties of biological specimens, even for in-vivo measurements. In this framework, a well-known approach consists in the measurement of radiation emitted by fluorescence, typically induced by inserting fluorescent proteins inside the samples. It is also of great interest to acquire three-dimensional images of such specimens, and this can be achieved by exploiting light sheet fluorescence microscopy: this technique provides optical sectioning by sequentially illuminating different slices of the sample and stacking them to generate the full volume, ensuring at the same time low photobleaching and phototoxicity. In addition, this approach can be combined with image restoration techniques to reduce blurring and noise, thus increasing image quality. However, when imaging 3D thick samples, deep regions suffer from loss of information due to attenuation and scattering phenomena caused by the propagation of light through dense tissues. In this work I show the advantages given by multiview imaging to fully retrieve 3D reconstructions of biological samples with a few millimetres dimension. The specimen is measured from different orientations, merging all the acquisitions to determine a single final volume containing high-quality 3D reconstructions. I improved a setup for light sheet microscopy by inserting optical and mechanical elements to remove shadowing artifacts generated by absorption centres, to tune the illumination beam and to mechanically rotate the sample between following acquisitions. Multiview imaging is achieved by vertically mounting the specimen between the objectives and rotating it to acquire different volumes. This approach requires fine alignment to merge all the measured images into the final volume, and it is desirable to deconvolve it to further enhance image quality. For this purpose, a recently published iterative algorithm for deconvolution from autocorrelation is applied: it provides efficient processing of multiview acquisitions, by exploiting intrinsic alignment ensured by autocorrelation properties and removing blurring and noise by means of deconvolution. In addition, I implemented a blind approach for this method, demonstrating that, starting from partial or absent knowledge of the blurring mechanism, it is possible to improve measurement quality and simultaneously retrieve information about the optical system. Re-sults reported in this work show how to achieve high-quality 3D imaging of a transparent sample of a few millimetres, merging the advantages given by multiple acquisitions and by effective image processing techniques.
La microscopia ottica è uno strumento per lo studio delle proprietà microscopiche di campioni biologici, ed è largamente impiegata anche per misure in-vivo. In questo ambito, un noto approccio è la microscopia di fluorescenza, che misura la radiazione emessa dal campione in seguito ad assorbimento di luce, tipicamente indotta tramite l’inserzione di proteine fluorescenti. Inoltre, è di grande interesse acquisire immagini tridimensionali di specie biologiche, che possono essere ottenute grazie alla microscopia light sheet: questa tecnica garantisce sezionamento ottico illuminando in modo sequenziale singoli piani all’interno del campione, poi uniti a formare il volume completo, e riduce notevolmente foto-tossicità e photo-bleaching. Successivamente, tecniche di elaborazione digitale delle immagini possono essere applicate alle misure per ridurre sfocatura e rumore, incrementando la qualità delle acquisizioni. Tuttavia, considerando campioni tridimensionali relativamente spessi, regioni più profonde soffrono di perdita di segnale a causa di fenomeni di assorbimento e scattering durante la propagazione della luce all’interno dei tessuti. In questo lavoro di tesi dimostro i vantaggi forniti da misure multiview per ricostruire immagini 3D di campioni biologici con dimensioni di pochi millimetri: essi vengono misurati da diverse orientazioni e tutte le acquisizioni sono successivamente fuse per determinare un singolo volume finale contenente una ricostruzione 3D di alta qualità. In particolare, ho perfezionato un setup per microscopia light sheet inserendo elementi ottici e meccanici per rimuovere ombre generate da centri di assorbimento, per variare e adattare il fascio di illuminazione tra diverse misure, e per ruotare il campione con un controllo meccanico durante misure successive. Immagini multiview sono ottenute montando il campione verticalmente tra gli obiettivi e ruotandolo per acquisire differenti volumi. Questo approccio richiede un processo accurato di allineamento per unire tutte le viste, inoltre per aumentare ulteriormente la qualità dell’immagine è conveniente applicare metodi numerici di deconvoluzione. A tal proposito, ho applicato un algoritmo di recente pubblicazione che unisce i processi di deconvoluzione e de-autocorrelazione, fornendo un’elaborazione efficiente delle acquisizioni multiview grazie all’allineamento intrinseco delle immagini nello spazio delle autocorrelazioni e rimuovendo sfocatura e rumore grazie alla deconvoluzione. Inoltre, ho implementato un approccio blind per tale algoritmo, dimostrando che, partendo da una conoscenza parziale o nulla riguardo al processo di degradazione della misura, è possibile migliorare la qualità della ricostruzione e allo stesso tempo recuperare informazioni sul sistema ottico. I risultati riportati in questa tesi mostrano come misurare campioni trasparenti grandi pochi millimetri e ottenere delle immagini tridimensionali di alta qualità, sfruttando l’acquisizione di volumi misurati a differenti orientazioni e applicando efficaci metodi di elaborazione digitale delle immagini.
Multiview light sheet fluorescence microscopy and autocorrelation-based methods for image restoration
CORBETTA, ELENA
2019/2020
Abstract
Optical microscopy is a powerful tool employed in the field of biology to investigate the microscopic properties of biological specimens, even for in-vivo measurements. In this framework, a well-known approach consists in the measurement of radiation emitted by fluorescence, typically induced by inserting fluorescent proteins inside the samples. It is also of great interest to acquire three-dimensional images of such specimens, and this can be achieved by exploiting light sheet fluorescence microscopy: this technique provides optical sectioning by sequentially illuminating different slices of the sample and stacking them to generate the full volume, ensuring at the same time low photobleaching and phototoxicity. In addition, this approach can be combined with image restoration techniques to reduce blurring and noise, thus increasing image quality. However, when imaging 3D thick samples, deep regions suffer from loss of information due to attenuation and scattering phenomena caused by the propagation of light through dense tissues. In this work I show the advantages given by multiview imaging to fully retrieve 3D reconstructions of biological samples with a few millimetres dimension. The specimen is measured from different orientations, merging all the acquisitions to determine a single final volume containing high-quality 3D reconstructions. I improved a setup for light sheet microscopy by inserting optical and mechanical elements to remove shadowing artifacts generated by absorption centres, to tune the illumination beam and to mechanically rotate the sample between following acquisitions. Multiview imaging is achieved by vertically mounting the specimen between the objectives and rotating it to acquire different volumes. This approach requires fine alignment to merge all the measured images into the final volume, and it is desirable to deconvolve it to further enhance image quality. For this purpose, a recently published iterative algorithm for deconvolution from autocorrelation is applied: it provides efficient processing of multiview acquisitions, by exploiting intrinsic alignment ensured by autocorrelation properties and removing blurring and noise by means of deconvolution. In addition, I implemented a blind approach for this method, demonstrating that, starting from partial or absent knowledge of the blurring mechanism, it is possible to improve measurement quality and simultaneously retrieve information about the optical system. Re-sults reported in this work show how to achieve high-quality 3D imaging of a transparent sample of a few millimetres, merging the advantages given by multiple acquisitions and by effective image processing techniques.| File | Dimensione | Formato | |
|---|---|---|---|
|
MasterThesis_Elena_Corbetta.pdf
non accessibile
Dimensione
3.51 MB
Formato
Adobe PDF
|
3.51 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/175682