Search

Immunocyte infiltration and cytotoxicity play critical roles in both inflammation and immunotherapy. However, current cancer immunotherapy screening methods overlook the capacity of the T cells to penetrate the tumor stroma, thereby significantly limiting the development of effective treatments for solid tumors. Here, we present an automated high-throughput microfluidic platform for simultaneous tracking of the dynamics of T cell infiltration and cytotoxicity within the 3D tumor cultures with a tunable stromal makeup. By recourse to a clinical tumor-infiltrating lymphocyte (TIL) score analyzer, which is based on a clinical data-driven deep learning method, our platform can evaluate the efficacy of each treatment based on the scoring of T cell infiltration patterns. By screening a drug library using this technology, we identified an epigenetic drug (lysine-specific histone demethylase 1 inhibitor, LSD1i) that effectively promoted T cell tumor infiltration and enhanced treatment efficacy in combination with an immune checkpoint inhibitor (anti-PD1) in vivo. We demonstrated an automated system and strategy for screening immunocyte-solid tumor interactions, enabling the discovery of immuno- and combination therapies.

Aero-turbines usually work in various non-stationary and harsh operating environments. Its blades easily appear early crack faults in long-time operation, and it is difficult to dig out discriminant features of early crack. To solve these problems, in the paper, a novel intelligent approach for early crack diagnosis of turbine blades using three-dimensional blade tip clearance is presented. In order to improve feature learning ability to obtain better generalization ability, the paper firstly develops a novel deep learning method based on deep belief networks (DBNs). Considering the fact that the feature degradation easily occurs in deeper layers because of the change of the distribution in each layer's outputs with the increase of the layers, it is hard to decide which layer to learn features is useful for fault diagnosis. Accordingly, in the pre-training process, the global back-reconstruction (GBR) mechanism is introduced into DBNs to optimize the feature learning ability. The GBR mechanism can be realized between the input layer and hidden layers by “shortcut connection”, and the layer to learn more discriminant features can be determined automatically without prior knowledge. Moreover, due to three-dimensional blade tip clearance (3-DBTC) acquired from three different directions of turbine blades contains much more useful crack failure multiscale information, it is suitable to be used as an input from which to extract multiscale discriminant features. Eventually, in the supervised training, the softmax regression model is employed to classify the health conditions of turbine blades using these sensitive features learned from 3-DBTC. The experimental results show that the proposed method can effectively identify the crack of turbine blades with fairly high diagnostic accuracies and significantly outperform other methods considered in the paper.

We propose a general learning based framework for solving nonsmooth and nonconvex image reconstruction problems. We model the regularization function as the composition of the $l_{2,1}$ norm and a smooth but nonconvex feature mapping parametrized as a deep convolutional neural network. We develop a provably convergent descent-type algorithm to solve the nonsmooth nonconvex minimization problem by leveraging the Nesterov's smoothing technique and the idea of residual learning, and learn the network parameters such that the outputs of the algorithm match the references in training data. Our method is versatile as one can employ various modern network structures into the regularization, and the resulting network inherits the guaranteed convergence of the algorithm. We also show that the proposed network is parameter-efficient and its performance compares favorably to the state-of-the-art methods in a variety of image reconstruction problems in practice.