Borrelli, Francesca (2023) Advanced Computational Methodologies in Tomographic Phase Imaging Flow Cytometry. [Tesi di dottorato]

Anteprima

Testo Borrelli_Francesca_36.pdf Download (20MB) | Anteprima

Abstract

Single-cell analysis will constitute the future of precision medicine and early diagnosis. To date, the gold-standard cytometric technique is Fluorescence Imaging Flow Cytometry (FIFC), which allows to record 2D images of stained single cells as they flow through a measuring device, granting an elevate throughput. Since FIFC can provide large, informative image datasets, the combination with Artificial Intelligence (AI) - based analysis tools can allow a fast, automatic, and objective cell phenotyping. However, the staining process can affect in different ways the cells under investigation. In fact, contrast agents can disturb the normal function of living cells (cytotoxicity) and lasers used in FIFC can be toxic for them as well (phototoxicity). Conversely, label-free optical imaging techniques avoid cells staining, thus becoming the most promising solution to overcome the intrinsic limitations of the FIFC. Among label-free microscopies, Tomographic Phase Microscopy (TPM) represents the highest informative imaging strategy as it allows to reconstruct the volumetric distribution of the Refractive Index (RI) at the single-cell level. The cellular RI is a key biophysical parameter proven to be an effective descriptor of cellular heterogeneity and physiopathology. For the first time in 2017, TPM has been demonstrated in Flow Cytometry (FC) mode, paving the way to the Tomographic Phase Imaging Flow Cytometry (TPIFC) technology. In TPIFC, digital holograms of single cells are recorded in continuous flow while the samples rotate in microfluidic environment. As it allows to conjugate the high throughput of FC technology and the availability of the full 3D RI information, the TPIFC tool is expected to create a breakthrough in the cell biology studies and in the clinical practice. Nowadays, the research and commercial trend is to combine TPIFC technology with Lab on Chip (LOC) devices to go towards telehealth applications. The desired LOC devices will be compact, cheap and characterised by inexpensive components and acceptable computational effort.This creates a challenge for TPIFC, going towards the paradigm of Computational Tomography, according to which the complexity will be shifted from the optical and hardware point of view to the numerical one. Accordingly, advanced computational methodologies are needed to achieve the above goal. In addition, to keep the elevate throughput of FC, fast tomographic approaches are sought after, able to reconstruct with a reduced number of acquisitions. Therefore, several computational strategies are developed in this Ph.D. Thesis for transferring the original proof of concept of TPIFC into a concrete, LOC-oriented technology for the single-cell analysis. In particular, powerful numerical approaches to reconstruct with reduced number of measurements and with low inference time are developed. Moreover, AI models for phenotyping cancer cells are discussed. Finally, a preliminary theoretical work is conducted for the higher-order description of the scattering process, to improve the tomographic imaging description and inversion approaches. In the near future, the attained results are expected to contribute in providing a solution to the challenging topic of the label-free single-cell analysis, for example focusing on early diagnosis of cancer and the development of personalised therapies by means of blood tests.

Downloads

Actions (login required)

Modifica documento