Dual-Angle Interferometric Scattering Microscopy for Optical Multiparametric Particle Characterization.

Traditional single-nanoparticle sizing using optical microscopy techniques assesses size via the diffusion constant, which requires suspended particles to be in a medium of known viscosity. However, these assumptions are typically not fulfilled in complex natural sample environments. Here, we introduce dual-angle interferometric scattering microscopy (DAISY), enabling optical quantification of both size and polarizability of individual nanoparticles (radius <170 nm) without requiring a priori information regarding the surrounding media or super-resolution imaging. DAISY achieves this by combining the information contained in concurrently measured forward and backward scattering images through twilight off-axis holography and interferometric scattering (iSCAT). Going beyond particle size and polarizability, single-particle morphology can be deduced from the fact that the hydrodynamic radius relates to the outer particle radius, while the scattering-based size estimate depends on the internal mass distribution of the particles. We demonstrate this by differentiating biomolecular fractal aggregates from spherical particles in fetal bovine serum at the single-particle level.

Single-shot self-supervised object detection in microscopy.

Object detection is a fundamental task in digital microscopy, where machine learning has made great strides in overcoming the limitations of classical approaches. The training of state-of-the-art machine-learning methods almost universally relies on vast amounts of labeled experimental data or the ability to numerically simulate realistic datasets. However, experimental data are often challenging to label and cannot be easily reproduced numerically. Here, we propose a deep-learning method, named LodeSTAR (Localization and detection from Symmetries, Translations And Rotations), that learns to detect microscopic objects with sub-pixel accuracy from a single unlabeled experimental image by exploiting the inherent roto-translational symmetries of this task. We demonstrate that LodeSTAR outperforms traditional methods in terms of accuracy, also when analyzing challenging experimental data containing densely packed cells or noisy backgrounds. Furthermore, by exploiting additional symmetries we show that LodeSTAR can measure other properties, e.g., vertical position and polarizability in holographic microscopy.

Convolutional neural networks for segmentation of FIB-SEM nanotomography data from porous polymer films for controlled drug release.

Phase-separated polymer films are commonly used as coatings around pharmaceutical oral dosage forms (tablets or pellets) to facilitate controlled drug release. A typical choice is to use ethyl cellulose and hydroxypropyl cellulose (EC/HPC) polymer blends. When an EC/HPC film is in contact with water, the leaching out of the water-soluble HPC phase produces an EC film with a porous network through which the drug is transported. The drug release can be tailored by controlling the structure of this porous network. Imaging and characterization of such EC porous films facilitates understanding of how to control and tailor film formation and ultimately drug release. Combined focused ion beam and scanning electron microscope (FIB-SEM) tomography is a well-established technique for high-resolution imaging, and suitable for this application. However, for segmenting image data, in this case to correctly identify the porous network, FIB-SEM is a challenging technique to work with. In this work, we implement convolutional neural networks for segmentation of FIB-SEM image data. The data are acquired from three EC porous films where the HPC phases have been leached out. The three data sets have varying porosities in a range of interest for controlled drug release applications. We demonstrate very good agreement with manual segmentations. In particular, we demonstrate an improvement in comparison to previous work on the same data sets that utilized a random forest classifier trained on Gaussian scale-space features. Finally, we facilitate further development of FIB-SEM segmentation methods by making the data and software used open access.

Drug release from pharmaceutical tablets or pellets is often controlled by applying a phase-separated polymer film coating. Ethyl cellulose and hydroxypropyl cellulose (EC/HPC) polymer blends are commonly used. The HPC phase leaches out when in contact with water and the result is a porous EC matrix coating, with mass transport properties that can be controlled by tailoring the structure of the porous network. High-resolution 3D imaging is necessary to characterize such materials, and the resolution of e.g. X-ray computed tomography is simply insufficient. Combined focused ion beam and scanning electron microscope (FIB-SEM) tomography on the other hand is a suitable technique, but segmentation of FIB-SEM data, in this case to separate the solid matrix and the porous network, is challenging. In this work, we develop a method for segmentation of FIB-SEM image data acquired from three different EC porous films where the HPC phases have been leached out. The segmentation is based on convolutional neural networks (CNNs). CNNs is a well-established machine learning paradigm and has demonstrated state-of-the-art performance in many image analysis and segmentation tasks. CNNs are inspired from biological processes in the visual cortex and act similarly, at least conceptually. In contrast to most conventional machine learning algorithms, CNNs learn by themselves which features to extract from the images. The features are extracted at different spatial scales and may constitute e.g. edge and contrast detectors. These features are subsequently used for classification. In this work, CNNs are used for image segmentation. The goal is to identify which regions in the images that contain either pore (empty space) or solid (material), hence a binary classification task. For the CNN to learn how to perform such a task, a ground truth is needed. This is achieved by letting an expert manually segment parts of the data. This is a very time-consuming endeavor, hence only a small random subset of the full dataset is manually segmented. The CNN is trained for the task using the manually segmented data, after which automatic segmentation of the full dataset is performed. We obtain very good agreement with manual segmentations in terms of accuracy and porosity, and a clear improvement in comparison to an earlier developed random forest classifier trained on Gaussian scale-space features on the same data. The development of accurate segmentation methods is a crucial step toward better understanding and tailoring of coatings for controlled drug release.