Digital holographic microscopy (DHM) is a well-known powerful method allowing both the amplitude and phase of a specimen to be simultaneously observed. In order to obtain a reconstructed image from a hologram, numerous calculations for the Fresnel diffraction are required. The Fresnel diffraction can be accelerated by the FFT (Fast Fourier Transform) algorithm. However, real-time reconstruction from a hologram is difficult even if we use a recent central processing unit (CPU) to calculate the Fresnel diffraction by the FFT algorithm. In this paper, we describe a real-time DHM system using a graphic processing unit (GPU) with many stream processors, which allows use as a highly parallel processor. The computational speed of the Fresnel diffraction using the GPU is faster than that of recent CPUs. The real-time DHM system can obtain reconstructed images from holograms whose size is 512 x 512 grids in 24 frames per second.

No abstract

This paper proposes a target vector modification method for the all-transfer deep learning (ATDL) method. Deep neural networks (DNNs) have been used widely in many applications; however, the DNN has been known to be problematic when large amounts of training data are not available. Transfer learning can provide a solution to this problem. Previous methods regularize all layers, including the output layer, by estimating the relation vectors, which are then used instead of one-hot target vectors of the target domain. These vectors are estimated by averaging the target domain data of each target domain label in the output space. This method improves the classification performance, but it does not consider the relation between the relation vectors. From this point of view, we propose a relation vector modification based on constrained pairwise repulsive forces. High pairwise repulsive forces provide large distances between the relation vectors. In addition, the risk of divergence is mitigated by the constraint based on distributions of the output vectors of the target domain data. We apply our method to two simulation experiments and a disease classification using two-dimensional electrophoresis images. The experimental results show that reusing all layers through our estimation method is effective, especially for a significantly small number of the target domain data.

We have developed a high-resolution chiral derivatization-LC method for the determination of D-and L-amino acids using 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMT-MM) as an enantioseparation enhancer. In this method, D-and L-amino acids were derivatized with (S)-(-)-1-(naphthyl)ethylamine (NEA) in the presence of DMT-MM as a condensing agent. These conditions resulted in the amino group of the D-and L-amino acid-NEA derivative being introduced in the triazine unit of DMT-MM, producing sterically bulky diastereoisomers. The resulting derivatives of the D-and L-amino acids could be discriminated and successfully enantioseparated through reversed-phase LC. This enantioseparation enhancement effect was demonstrated when using six other commercially available chiral amines as derivatization reagents. Since triazine modification possesses high ESI-responsivity, they showed higher sensitivities than non-triazine derivatives. The developed derivatization method was successfully applied to highly sensitive quantification of D-and L-amino acids by combining LC-MS/MS analysis.

An accuracy of ≥98% was achieved in sepsis diagnosis using serum samples from 30 sepsis patients and 68 healthy individuals and a high-performance two-dimensional polyacrylamide gel electrophoresis (HP-2D-PAGE) method developed here with deep learning and transfer learning algorithms. In this method, small-scale target domain data, which are collected to achieve our objective, are inputted directly into a model constructed with source domain data which are collected from a different domain from the target; target vectors are estimated with the outputted target domain data and applied to refine the model. Recognition performance of small-scale data is improved by reusing all layers, including the output layers of the neural network. Proteomics is generally considered the ultimate bio-diagnostic technique and provides extremely high information density in its two-dimensional electrophoresis images, but extracting the data has posed a basic problem. The present study is expected to solve that problem and will be an important breakthrough for practical utilization and future perspectives of proteomics in clinics after evaluation in clinical settings.