Lingdao Sha

Background: Tumor programmed death-ligand 1 (PD-L1) status is useful in determining which patients may benefit from programmed death-1 (PD-1)/PD-L1 inhibitors. However, little is known about the association between PD-L1 status and tumor histopathological patterns. Using deep learning, we predicted PD-L1 status from hematoxylin and eosin (H and E) whole-slide images (WSIs) of nonsmall cell lung cancer (NSCLC) tumor samples. Materials and Methods: One hundred and thirty NSCLC patients were randomly assigned to training (n = 48) or test (n = 82) cohorts. A pair of H and E and PD-L1-immunostained WSIs was obtained for each patient. A pathologist annotated PD-L1 positive and negative tumor regions on the training samples using immunostained WSIs for reference. From the H and E WSIs, over 145,000 training tiles were generated and used to train a multi-field-of-view deep learning model with a residual neural network backbone. Results: The trained model accurately predicted tumor PD-L1 status on the held-out test cohort of H and E WSIs, which was balanced for PD-L1 status (area under the receiver operating characteristic curve [AUC] =0.80, P<< 0.01). The model remained effective over a range of PD-L1 cutoff thresholds (AUC = 0.67-0.81, P ≤ 0.01) and when different proportions of the labels were randomly shuffled to simulate interpathologist disagreement (AUC = 0.63-0.77, P ≤ 0.03). Conclusions: A robust deep learning model was developed to predict tumor PD-L1 status from H and E WSIs in NSCLC. These results suggest that PD-L1 expression is correlated with the morphological features of the tumor microenvironment.

CONTEXT: Color normalization techniques for histology have not been empirically tested for their utility for computational pathology pipelines. AIMS: We compared two contemporary techniques for achieving a common intermediate goal - epithelial-stromal classification. SETTINGS AND DESIGN: Expert-annotated regions of epithelium and stroma were treated as ground truth for comparing classifiers on original and color-normalized images. MATERIALS AND METHODS: Epithelial and stromal regions were annotated on thirty diverse-appearing H and E stained prostate cancer tissue microarray cores. Corresponding sets of thirty images each were generated using the two color normalization techniques. Color metrics were compared for original and color-normalized images. Separate epithelial-stromal classifiers were trained and compared on test images. Main analyses were conducted using a multiresolution segmentation (MRS) approach; comparative analyses using two other classification approaches (convolutional neural network [CNN], Wndchrm) were also performed. STATISTICAL ANALYSIS: For the main MRS method, which relied on classification of super-pixels, the number of variables used was reduced using backward elimination without compromising accuracy, and test - area under the curves (AUCs) were compared for original and normalized images. For CNN and Wndchrm, pixel classification test-AUCs were compared. RESULTS: Khan method reduced color saturation while Vahadane reduced hue variance. Super-pixel-level test-AUC for MRS was 0.010-0.025 (95% confidence interval limits ± 0.004) higher for the two normalized image sets compared to the original in the 10-80 variable range. Improvement in pixel classification accuracy was also observed for CNN and Wndchrm for color-normalized images. CONCLUSIONS: Color normalization can give a small incremental benefit when a super-pixel-based classification method is used with features that perform implicit color normalization while the gain is higher for patch-based classification methods for classifying epithelium versus stroma.

Sparse coding is widely used in image denoising, deblurring, clustering, and classification. However, most existing approaches to sparse coding failed to consider the fact that high dimensional data naturally reside on geometrical structure of the data space. It has been shown that geometric information of the data is important for both inversion and discrimination. In this paper, we proposed a generalized framework for image restoration and representation by combining sparse coding and graph based algorithms. In image denoising and deblurring problems, an image is first decomposed into cartoon layer (piecewise-smooth contents) and texture layer (textures and sharp edges) using morphological component analysis (MCA); then optimal graph Laplacian regularizer (OGLR) algorithm and simultaneous sparse coding with Gaussian scale mixture prior (SSC-GSM) algorithm are applied to cartoon layer and texture layer, respectively; final restored image is generated by aggregating the outcomes from two algorithms. The proposed hybrid image restoration algorithm outperforms state-of-the-art image denoising algorithms, such as BM3D on natural images, measured in PSNR, and performs comparatively in image deblurring. In image clustering and classification problems, we convert our generalized framework into a novel dual graph regularized sparse coding method to transform the nonlinear data space and feature space into linear space, two efficient optimization algorithms are provided for the numerical implementation. The experimental results show that our generalized graph Laplacian and sparse coding framework performs competitively with popular denoising, deblurring, clustering, and classification methods.

Computer based automatic medical image processing and quantification are becoming popular in digital pathology. However, preparation of histology slides can vary widely due to differences in staining equipment, procedures and reagents, which can reduce the accuracy of algorithms that analyze their color and texture information. To re- duce the unwanted color variations, various supervised and unsupervised color normalization methods have been proposed. Compared with supervised color normalization methods, unsupervised color normalization methods have advantages of time and cost efficient and universal applicability. Most of the unsupervised color normaliza- tion methods for histology are based on stain separation. Based on the fact that stain concentration cannot be negative and different parts of the tissue absorb different stains, nonnegative matrix factorization (NMF), and particular its sparse version (SNMF), are good candidates for stain separation. However, most of the existing unsupervised color normalization method like PCA, ICA, NMF and SNMF fail to consider important information about sparse manifolds that its pixels occupy, which could potentially result in loss of texture information during color normalization. Manifold learning methods like Graph Laplacian have proven to be very effective in interpreting high-dimensional data. In this paper, we propose a novel unsupervised stain separation method called graph regularized sparse nonnegative matrix factorization (GSNMF). By considering the sparse prior of stain concentration together with manifold information from high-dimensional image data, our method shows better performance in stain color deconvolution than existing unsupervised color deconvolution methods, especially in keeping connected texture information. To utilized the texture information, we construct a nearest neighbor graph between pixels within a spatial area of an image based on their distances using heat kernal in <i>l</i>&alpha;&beta; space. The representation of a pixel in the stain density space is constrained to follow the feature distance of the pixel to pixels in the neighborhood graph. Utilizing color matrix transfer method with the stain concentrations found using our GSNMF method, the color normalization performance was also better than existing methods.

The goal for video game AI (artificial intelligence) is to generate AI that is at appropriate challenge level. Most existing game AI is implemented by FSM (Finite State Machine) which has drawbacks in the three respects: requirement of designer's intensive participation; can't adjust strategies or difficulty dynamically; no planning and looking forward. Contribution of this paper is to propose DDA (dynamic difficulty adjustment) as an approach to create appropriate challenge level game opponent. During the research, the prey and predator genre game of Dead-End is used as test-bed to prove the proposed theory. Based on the Dead-End test-bed, I proposed two kinds of DDA which are DDA by “time-constrained-CI” and DDA by “knowledge-based-time-constrained-CI”. As the latter is based on knowledge, it is more computational resource efficient than the former and thus more applicable for multi-player online games, while the former is only applicable for the standalone PC game.