Deep learning classification of lipid droplets in quantitative phase images

Abstract

We report the application of supervised machine learning to the automated classification of lipid droplets in label-free, quantitative-phase images. By comparing various machine learning methods commonly used in biomedical imaging and remote sensing, we found convolutional neural networks to outperform others, both quantitatively and qualitatively. We describe our imaging approach, all implemented machine learning methods, and their performance with respect to computational efficiency, required training resources, and relative method performance measured across multiple metrics. Overall, our results indicate that quantitative-phase imaging coupled to machine learning enables accurate lipid droplet classification in single living cells. As such, the present paradigm presents an excellent alternative of the more common fluorescent and Raman imaging modalities by enabling label-free, ultra-low phototoxicity, and deeper insight into the thermodynamics of metabolism of single cells.

Speeding up reconstruction of 3D tomograms in holographic flow cytometry via deep learning

“…In fact, from the QPMs, 2D labelfree features can be measured for diagnostic purposes. [5][6][7][8][9][10][11][12][13][14][15][16][17] As a consequence, a further way to validate the architecture we propose consists of checking whether the QPM outputs lead to the same features that would be measured from the corresponding QPM targets. To this aim, the 2000 cells belonging to the test set have been segmented from the background within the QPMs.…”

“…[1][2][3] Thanks to these features, DH has been successfully employed in a variety of biomedicine applications, 4 including cancer cell identification and characterization, [5][6][7] diagnostics of blood diseases, [8][9][10][11] inflammations 12 and infectious diseases, [13][14][15] study of cell motility and migration, 16 and marker-free detection of lipid droplets (LDs). 17 The possibility to probe biological samples from different directions leads to the full 3D label-free imaging achieved by holographic tomography technology, 18,19 which represents the leading edge of biological inspection at the single-cell level. The combination of compact holographic microscopy and flow cytometry allows the high-throughput screening of cells flowing in microfluidic channels, thus permitting biological specimens to be studied in their natural environment for point-of-care diagnostics at the lab-on-a-chip scale.…”

3D imaging lipidometry in single cell by in-flow holographic tomography

“…The RI values of LDs change based on the type of cell, the temperature, and the wavelength 43 . Segmenting intracellular organelles is always problematic in a label-free technique since an exogenous calibrated marker is missed.…”

Quantitative Phase Imaging: Recent Advances and Expanding Potential in Biomedicine

“…For example, machine learning can be used to identify lipid droplets in unlabeled QPI images. 337 A related machine learning approach, called PhaseStain, was developed for label-free staining of QPI images. 338 This method was extended for realtime staining and classification of sperm cells, 339 identification of cells from subcellular components, 340 and generation of pseudofluorescence images from label-free QPI data.…”