In flow manipulation and characterization of cancer cells by coherent computational microscopy

Liquid biopsy has shown remarkably promising in oncology for the early diagnosis of cancer through the detection of circulating biomarkers such as circulating tumor cells (CTCs). Recent evidences suggest that CTCs represent effective prognostic and predictive biomarkers to monitor/predict therapy efficacy in breast, colon and prostate cancers [1,2]. However, the frequency of CTCs in blood is approximately 1 to 10 cells per 10 mL of blood, which is as challenging as looking for a needle in a haystack. In microfluidics, Digital Holography (DH) has been shown to be a promising technique to characterize CTCs with the aim to detect them inside a heterogeneous liquid sample. DH is label-free, real-time and gives access to the complex amplitude of the object [3-6]. Thus, any classification approach based on the holographic signature can exploit a reach information content to take a decision. Moreover, the flexible refocusing capability of DH imaging allows to inspect an entire liquid volume with a single capture. This enables the high-throughput inspection of blood and other bodily fluids rapidly flowing inside microfluidic channels. In DH, the sample is probed from one single direction and the phase delay introduced by the sample in through transmission acts as a contrast agent. Hence, the optical thickness measurable by DH imaging is an integral information, i.e. the sum of all the contributions experienced by the coherent light during its passage through the sample. In order to decouple the refractive index from the physical thickness and to resolve its distribution along the optical axis, tomography exploits multiple recordings, probing the sample from different angles and combining the corresponding phase-contrast maps [7,8]. Various schemes have been proposed to minimize the number of sampling angles and to make the recording stage faster in order to match the requirements and time constraints imposed by real biological problems. Here we show the recent advances of in-flow holographic tomography, which exploits a controlled induced rotation of the sample inside the microfluidic channel to probe it from different view angles with no mechanical rotation of the source beam [8,9]. We introduce an effective algorithm to recover from the recorded phase maps the set of angles required as input of the optical projection tomography algorithm [7-9]. We show the application of holographic flow tomography to the characterization of different cancer cells [10], namely breast cancer cells, ovarian cancer cells and neuroblastoma. We also discuss different possibilities of Lab-on-a-Chip design and flow engineering that allow us to induce controlled rotations while maintaining the high-throughput nature of DH microscopy [9,11,12]. In the next future, the large amount of data obtainable by this approach will be used to train a neural network devoted to classify CTCs, distinguishing them from the other components of a blood stream.