Kazuyuki Nakamura

Long-term polarization-sensitive micro-pulse lidar (PMPL) observations provide valuable information on aerosol transport over coastal Antarctica. Still, the interpretation of aerosolenhanced layers has largely relied on manual inspection of quick-look images. Here we propose an objective image-based framework for identifying aerosol transport regimes from PMPL observations at Syowa Station, East Antarctica, during 2016–2022. Aerosol-enhanced layers were first detected using the backscatter ratio, its temporal variability, and the particle depolarization ratio. Image features were then extracted from PMPL quick-look images using a pre-trained ResNet50 convolutional neural network, standardized, reduced in dimensionality by principal component analysis, and classified by k-means clustering. A total of 254 aerosol events were identified and classified into two transport regimes with distinct vertical structures: one characterized by vertically extensive aerosol layers and the other by shallow, near-surface layers. Independent composite analyses of ERA5 500-hPa geopotential height anomalies indicate that the two regimes occur under different synoptic environments. These results suggest that PMPL observations alone contain sufficient information to objectively identify physically meaningful aerosol transport regimes without using meteorological information in the classification. The framework provides a reproducible tool for long-term analyses of Antarctic aerosol transport.

Anatomical landmark localization is gaining attention to ease the burden on physicians. Focusing on aortic root landmark localization, the three hinge points of the aortic valve can reduce the burden by automatically determining the valve size required for transcatheter aortic valve implantation surgery. Existing methods for landmark prediction of the aortic root mainly use time-consuming two-step estimation methods. We propose a highly accurate one-step landmark localization method from even coarse images. The proposed method uses an optimal transport loss to break the trade-off between prediction precision and learning stability in conventional heatmap regression methods. We apply the proposed method to the 3D CT image dataset collected at Sendai Kousei Hospital and show that it significantly improves the estimation error over existing methods and other loss functions. Our code is available on GitHub.

Representation learning (RL) is a universal technique for deriving low-dimensional disentangled representations from high-dimensional observations, aiding a multitude of downstream tasks. RL has been extensively applied to various data types, including images and natural language. Here, we analyze molecular dynamics (MD) simulation data of biomolecules in terms of RL to obtain disentangled representations related to their conformational transitions. Currently, state-of-the-art RL techniques, which are mainly motivated by the variational principle, try to capture slow motions in the representation (latent) space. Here, we propose two methods based on alternative perspective on the disentanglement in the representation space. The methods introduce a simple prior that imposes temporal constraints in the representation space, serving as a regularization term to facilitate capturing disentangled representations of dynamics. The introduction of this simple prior aids in characterizing the conformational transitions of proteins. Indeed, comparison with other methods via the analysis of MD simulation trajectories for alanine dipeptide and chignolin validates that the proposed methods construct Markov state models (MSMs) whose implied time scales are comparable to state-of-the-art methods. By coarse-graining MSMs, we further show the methods aid to detect physically important interactions for conformational transitions. Overall, our methods provide good representations of complex biomolecular dynamics for downstream tasks, allowing for better interpretations of conformational transitions.

Time series model inference can be divided into modeling and optimization. Sequential VAEs have been studied as a modeling technique. As an optimization technique, methods combining variational inference (VI) and sequential Monte Carlo (SMC) have been proposed; however, they have two drawbacks: less particle diversity and biased gradient estimators. This paper proposes Ensemble Kalman Variational Objective (EnKO), a VI framework with the ensemble Kalman filter, to infer latent time-series models. Our proposed method efficiently learns the time-series models because of its particle diversity and unbiased gradient estimators. We demonstrate that our EnKO outperforms previous SMC-based VI methods in the predictive ability for several synthetic and real-world data sets.

The Kalman filter has been shown to be the most powerful tool for estimating the states of a linear Gaussian system. With this method, an expectation maximization algorithm can be used to estimate the model's parameters. However, the algorithm is unable to function in real-time due to the enormous computation and memory costs for smoothing. We propose three methods by which to estimate the model's states and parameters in real-time. We have demonstrated that the method correctly estimates the states and the parameters for the damped oscillation model. In addition, by introducing localization and spatial uniformity to the proposed method, we have demonstrated that the methods effectively denoise and short-term forecast for high-dimensional spatio-temporal data. The results indicate that the proposed methods can be applied in various fields, such as weather forecasting and vector field analysis.

Mechanical forces are critical for the emergence of diverse three-dimensional morphologies of multicellular systems. However, it remains unclear what kind of mechanical parameters at cellular level substantially contribute to tissue morphologies. This is largely due to technical limitations of live measurements of cellular forces. Here we developed a framework for inferring and modeling mechanical forces of cell–cell interactions. First, by analogy to coarse-grained models in molecular and colloidal sciences, we approximated cells as particles, where mean forces (i.e. effective forces) of pairwise cell–cell interactions are considered. Then, the forces were statistically inferred by fitting the mathematical model to cell tracking data. This method was validated by using synthetic cell tracking data resembling variousin vivosituations. Application of our method to the cells in the early embryos of mice and the nematodeCaenorhabditis elegansrevealed that cell–cell interaction forces can be written as a pairwise potential energy in a manner dependent on cell–cell distances. Importantly, the profiles of the pairwise potentials were quantitatively different among species and embryonic stages, and the quantitative differences correctly described the differences of their morphological features such as spherical vs. distorted cell aggregates, and tightly vs. non-tightly assembled aggregates. We conclude that the effective pairwise potential of cell–cell interactions is a live measurable parameter whose quantitative differences can be a parameter describing three-dimensional tissue morphologies.Author summary: Emergence of diverse three-dimensional morphologies of multicellular organisms is one of the most intriguing phenomena in nature. Due to the complex situations in living systems (e.g. a lot of genes are involved in morphogenesis.), a model for describing the emergent properties of multicellular systems has not been established. To approach this issue, approximation of the complex situations to limited numbers of parameters is required. Here, we searched for mechanical parameters for describing morphologies. We developed a statistical method for inferring mechanical potential energy of cell–cell interactions in three-dimensional tissues; the mechanical potential is an approximation of various mechanical components such as cell–cell adhesive forces, cell surface tensions, etc. Then, we showed that the quantitative differences in the potential is sufficient to reproduce basic three-dimensional morphologies observed during the mouse andC. elegansearly embryogenesis, revealing a direct link between cellular level mechanical parameters and three-dimensional morphologies. Our framework provides a noninvasive tool for measuring spatiotemporal cellular forces, which would be useful for studying morphogenesis of larger tissues including organs and their regenerative therapy.

Large aperture reflectors can have high radio frequency (RF) gain due to its large aperture and can be used to make flexible and appropriate services with recent digital processing technologies. Despite the performance of large aperture reflectors, most commercial geostationary communications satellites use up to 2.5 m aperture reflectors for high speed and fixed beam communication services due to rocket fairing size restrictions. To improve this situation, the recent discussions held by the Ministry of Internal Affairs and Communications Japan have concluded that large aperture reflector antennas for Ka-band are one of the key technologies to realize for the next generation of geostationary communications satellites. To this end, JAXA, or Japan Aerospace Exploration Agency, started to investigate the realization of 5 m aperture large deployable reflector of Ku/Ka-band for geostationary communication satellites. The reflector is a fan-fold deployable reflector and composed of a fan-fold deployable structure, a thin-plate network, and a metal mesh. The fan-fold deployable structure is designed as a Japanese fan, or “sensu,” and can be stowed in the manner of a folded fan. A rhombus lattice thin-plate network forms the parabolic surface. This chapter discusses the design concept of a fan-fold reflector and its applications.