Light Sheet Fluorescence Microscopy (LSFM) also known as Selective Plane Illumination Microscopy (SPIM), is a fluorescence microscopy technique which, from its first publication in 2004, has been increasingly used in biological applications, from developmental biology to mounted tissues analysis. As an alternative to LSFM, Selective Volume Illumination (SVI) microscopy has been recently developed with the goal of further increasing the acquisition rate. As is the case of LSFM, SVI uses a perpendicular illumination and detection. In this technique, rather than illuminating the sample in a single plane, the incoming excitation beam is modulated across a confined volume. On top of this, Light emitting Diodes (LED) are usually not considered a good light source for Light Sheet Fluorescence Microscopy because the lack of spatial coherence makes it difficult to tightly focus light along one direction. However, using LED in fluorescence microscopy brings advantages such as colour availability, low cost, and reduced presence of speckle patterns. In my PhD I demonstrated that, upon selectively illuminating a volume of the sample with LEDs, it is possible to volumetrically reconstruct it. The volumetric light modulation is possible thanks to a spatial light modulator, which is coupled to an illumination orthogonal to the detection axis. In this configuration, the modulation happens within the depth of field of the detection objective. After the acquisition of N patterns, an inversion problem is solved, resulting in the sample volumetric reconstruction. Furthermore, I illustrated how the use of patterned acquisition is compatible with Compressive Sensing, a signal analysis technique that reduces the number of modulation patterns to be acquired for a 3-dimensional reconstruction, compared to that given by the classical Nyquist-Shannon sampling criterion. This is achieved upon solving an ill-posed problem with further mathematical constraints which are related to the sample spatial features. The technique yields an accurate reconstruction of the sample anatomy even at significant compression ratios, up to compressed reconstruction in which only the 12.5% of the whole patterns set was used. Hence, the technique achieves higher volumetric acquisition rate, and eventually reduces photodamage on biological samples. In my PhD, I also demonstrated that the presented technique yields the absence of shadowing artifacts which are removed thanks to the broad spatial frequency support of an incoherent light source such as an LED. Finally, once I obtained a technique in which the volume of imaging was ultimately limited by the sensor dimensions in the thesis it is reported a computational strategy based on deconvolution to overcome the limited Depth of Field extension. This limitation in fact hinders the imaging of thicker samples. In the last part of my PhD, I extensively tested the cited strategies altogether, to assess for the overall achieved performances in samples like Danio rerio, a biological model in developmental and ecotoxicological studies and Mus musculus, i.e., lab mouse, imaged in different imaging media, both in vivo and ex vivo.
La microscopia a foglio di luce (LSFM), anche conosciuta come SPIM, è una tecnica di microscopia di fluorescenza che, dalla sua prima pubblicazione nel 2004, è stata sempre più usata per applicazioni biologiche, dalla biologia dello sviluppo all’ analisi di tessuti. Come alternativa alla LSFM, la microscopia a illuminazione selettiva di volume (SVIM), è stata recentemente sviluppata con l’obiettivo di aumentare ulteriormente la velocità di acquisizione. Come nel caso della LSFM, la SVIM usa un’illuminazione perpendicolare al ramo di acquisizione. In questa tecnica, piuttosto che illuminare il campione su di un singolo piano, il fascio di illuminazione è modulato su di un volume limitato. Inoltre, diodi a emissione di luce (LED), non sono di solito considerato come buone sorgenti per la microscopia a foglio di luce poiché loro mancanza di coerenza spaziale rende difficile focalizzare finemente la luce su di una singola dimensione. In ogni caso, l’uso di LED in microscopia di fluorescenza porta vantaggi come disponibilità cromatica, basso costo e presenza ridotta di pattern a macchie. Nel mio dottorato ho dimostrato come, dopo aver selettivamente illuminato una parte di campione con un LED, sia possibile ricostruirlo volumetricamente. La modulazione volumetrica della luce è possibile grazie ad un modulatore spaziale di luce, che è accoppiato con un’illuminazione perpendicolare all’asse di raccolta. In questa configurazione, la modulazione è contenuta nella profondità di campo dell’obiettivo di raccolta. Dopo aver acquisito N pattern, si risolve un problema inverso, che ha come risultato la ricostruzione volumetrica del campione. Oltretutto, Ho illustrato come l’impiego di pattern di acquisizione sia compatibile col Compressive Sensing (CS), una tecnica di analisi dei segnali che riduce il numero di pattern di modulazione da acquisire per una ricostruzione tridimensionale, rispettivamente al numero dato dal criterio classico di campionamento di Nyquist-Shannon. Ciò è ottenuto risolvendo un problema mal posto con ulteriori vincoli alla soluzione legati alle caratteristiche spaziali del campione. La tecnica apporta una ricostruzione accurata dell’anatomia del campione anche a compressioni significative, fino a ricostruzioni compresse in cui solo il 12.5% dei pattern totali è stato usato. Quindi la tecnica raggiunge velocità di acquisizione più elevate ed ottiene infine un ridotto danno al campione da esposizione alla luce. Nel mio PhD, ho anche dimostrato che la tecnica presentata mostra un’assenza di artefatti d’ombra, che sono rimossi grazie all’ampio supporto di frequenze spaziali di una sorgente di luce incoerente come un LED. Infine, una volta ottenuta una tecnica il cui volume di imaging fosse in ultimo limitato dalle dimensioni del sensore, nella tesi è riportato una strategia computazionale basata sulla deconvoluzione per superare la limitata estensione della profondità di campo. Questo limite, infatti, impedisce di esaminare campioni più spessi. Nell’ultima parte del mio dottorato, ho testato estensivamente nel complesso le sopracitate strategie, per quantificare le performance ottenute, in campioni come Danio rerio, un modello biologico per studi sullo sviluppo e sull’ecotossicologia, e come Mus musculus, topo da laboratorio, studiati in diversi mezzi di imaging, sia in vivo che ex vivo.
Compressive sensing in Light Sheet Fluorescence Microscopy
CALISESI, GIANMARIA
2020/2021
Abstract
Light Sheet Fluorescence Microscopy (LSFM) also known as Selective Plane Illumination Microscopy (SPIM), is a fluorescence microscopy technique which, from its first publication in 2004, has been increasingly used in biological applications, from developmental biology to mounted tissues analysis. As an alternative to LSFM, Selective Volume Illumination (SVI) microscopy has been recently developed with the goal of further increasing the acquisition rate. As is the case of LSFM, SVI uses a perpendicular illumination and detection. In this technique, rather than illuminating the sample in a single plane, the incoming excitation beam is modulated across a confined volume. On top of this, Light emitting Diodes (LED) are usually not considered a good light source for Light Sheet Fluorescence Microscopy because the lack of spatial coherence makes it difficult to tightly focus light along one direction. However, using LED in fluorescence microscopy brings advantages such as colour availability, low cost, and reduced presence of speckle patterns. In my PhD I demonstrated that, upon selectively illuminating a volume of the sample with LEDs, it is possible to volumetrically reconstruct it. The volumetric light modulation is possible thanks to a spatial light modulator, which is coupled to an illumination orthogonal to the detection axis. In this configuration, the modulation happens within the depth of field of the detection objective. After the acquisition of N patterns, an inversion problem is solved, resulting in the sample volumetric reconstruction. Furthermore, I illustrated how the use of patterned acquisition is compatible with Compressive Sensing, a signal analysis technique that reduces the number of modulation patterns to be acquired for a 3-dimensional reconstruction, compared to that given by the classical Nyquist-Shannon sampling criterion. This is achieved upon solving an ill-posed problem with further mathematical constraints which are related to the sample spatial features. The technique yields an accurate reconstruction of the sample anatomy even at significant compression ratios, up to compressed reconstruction in which only the 12.5% of the whole patterns set was used. Hence, the technique achieves higher volumetric acquisition rate, and eventually reduces photodamage on biological samples. In my PhD, I also demonstrated that the presented technique yields the absence of shadowing artifacts which are removed thanks to the broad spatial frequency support of an incoherent light source such as an LED. Finally, once I obtained a technique in which the volume of imaging was ultimately limited by the sensor dimensions in the thesis it is reported a computational strategy based on deconvolution to overcome the limited Depth of Field extension. This limitation in fact hinders the imaging of thicker samples. In the last part of my PhD, I extensively tested the cited strategies altogether, to assess for the overall achieved performances in samples like Danio rerio, a biological model in developmental and ecotoxicological studies and Mus musculus, i.e., lab mouse, imaged in different imaging media, both in vivo and ex vivo.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD Thesis_Calisesi.pdf
accessibile in internet per tutti
Descrizione: Final Thesis
Dimensione
2.94 MB
Formato
Adobe PDF
|
2.94 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/10589/177845