Search for: All records

We propose a novel Learned Alternating Minimization Algorithm (LAMA) for dual-domain sparse-view CT image reconstruction. LAMA is naturally induced by a variational model for CT reconstruction with learnable nonsmooth nonconvex regularizers, which are parameterized as composite functions of deep networks in both image and sinogram domains. To minimize the objective of the model, we incorporate the smoothing technique and residual learning architecture into the design of LAMA. We show that LAMA substantially reduces network complexity, improves memory efficiency and reconstruction accuracy, and is provably convergent for reliable reconstructions. Extensive numerical experiments demonstrate that LAMA outperforms existing methods by a wide margin on multiple benchmark CT datasets. 

Characterization of soluble protein aggregates provides valuable information for revealing mechanisms of protein aggregation process and assessing the activity and safety of protein therapeutics. However, the noncovalent interaction, the transient nature and higher degree of structural heterogeneity of the soluble aggregation system hinders precise characterization at the molecular level. Here, we describe methods using native mass spectrometry coupled with temperature-control electrospray ionization and size-exclusion chromatography to monitor the aggregation process and profile the aggregates in detail. 

By studying charge trapping in germanium detectors operating at temperatures below 10 K, we demonstrate for the first time that the formation of cluster dipole states from residual impurities is responsible for charge trapping. Two planar detectors with different impurity levels and types are used in this study. When drifting the localized charge carriers created by α particles from the top surface across a detector at a lower bias voltage, significant charge trapping is observed when compared to operating at a higher bias voltage. The amount of charge trapping shows a strong dependence on the type of charge carriers. Electrons are trapped more than holes in a p-type detector, while holes are trapped more than electrons in an n-type detector. When both electrons and holes are drifted simultaneously using the widespread charge carriers created by γ rays inside the detector, the amount of charge trapping shows no dependence on the polarity of bias voltage.