No abstract

The recent progress of computing, machine learning, and especially deep learning, for image recognition brings a meaningful effect for automatic detection of various diseases from chest X-ray images (CXRs). Here efficiency of lung segmentation and bone shadow exclusion techniques is demonstrated for analysis of 2D CXRs by deep learning approach to help radiologists identify suspicious lesions and nodules in lung cancer patients. Training and validation was performed on the original JSRT dataset (dataset #01), BSE-JSRT dataset, i.e. the same JSRT dataset, but without clavicle and rib shadows (dataset #02), original JSRT dataset after segmentation (dataset #03), and BSE-JSRT dataset after segmentation (dataset #04). The results demonstrate the high efficiency and usefulness of the considered pre-processing techniques in the simplified configuration even. The pre-processed dataset without bones (dataset #02) demonstrates the much better accuracy and loss results in comparison to the other pre-processed datasets after lung segmentation (datasets #02 and #03).

The results of chest X-ray (CXR) analysis of 2D images to get the statistically reliable predictions (availability of tuberculosis) by computer-aided diagnosis (CADx) on the basis of deep learning are presented. They demonstrate the efficiency of lung segmentation, lossless and lossy data augmentation for CADx of tuberculosis by deep convolutional neural network (CNN) applied to the small and not well-balanced dataset even. CNN demonstrates ability to train (despite overfitting) on the pre-processed dataset obtained after lung segmentation in contrast to the original not-segmented dataset. Lossless data augmentation of the segmented dataset leads to the lowest validation loss (without overfitting) and nearly the same accuracy (within the limits of standard deviation) in comparison to the original and other pre-processed datasets after lossy data augmentation. The additional limited lossy data augmentation results in the lower validation loss, but with a decrease of the validation accuracy. In conclusion, besides the more complex deep CNNs and bigger datasets, the better progress of CADx for the small and not well-balanced datasets even could be obtained by better segmentation, data augmentation, dataset stratification, and exclusion of non-evident outliers.

Background: Uterine cancer, also known as endometrial cancer, can seriously affect the female reproductive organs, and histopathological image analysis is the gold standard for diagnosing endometrial cancer. Computer-aided diagnosis (CADx) approaches based on traditional machine learning algorithms have been proposed to assist pathologists in interpreting histopathological images efficiently. However, due to the limited capability of modeling the complicated relationships between histopathological images and their interpretations, these CADx approaches often failed to achieve satisfying results. Methods: In this study, we developed a CADx approach using a convolutional neural network (CNN) and attention mechanisms, called HIENet. Because HIENet used the attention mechanisms and feature map visualization techniques, it can provide pathologists better interpretability of diagnoses by highlighting the histopathological correlations of local (pixel-level) image features to morphological characteristics of endometrial tissue. We then evaluated the classification performance of HIENet in ten-fold cross-validation on ~3,300 hematoxylin and eosin (H&E) images (collected from ~500 endometrial specimens from October 2017 to August 2018) and external validation on additional 200 H&E images (collected from 50 randomly-selected female patients during the first quarter of 2019). Results: In the ten-fold cross-validation process, the CADx approach achieved a 76.91 ± 1.17% (mean ± s. d.) classification accuracy for four classes of endometrial tissue, namely normal endometrium, endometrial polyp, endometrial hyperplasia, and endometrial adenocarcinoma. Also, HIENet achieved an area-under-the-curve (AUC) of 0.9579 ± 0.0103 with an 81.04 ± 3.87% sensitivity and 94.78 ± 0.87% specificity in a binary classification task that detected endometrioid adenocarcinoma ("Malignant"). Besides, in the external validation process, the CADx approach achieved an 84.50% accuracy in the four-class classification task, and it achieved an AUC of 0.9829 with a 77.97% (95% CI, 65.27%-87.71%) sensitivity and 100% (95% CI, 97.42%-100.00%) specificity. Moreover, positive predictive value (PPV) and negative predictive value (NPV) reached 100% (95% CI, 92.29%-100.00%) and 91.56% (95% CI, 86.00%-95.43%), respectively. The classification performance of HIENet can be further improved if directly trained as a binary classification model (also known as a binary classifier). Conclusion: The proposed CADx approach, HIENet, outperformed three human experts and four end-to-end CNN-based classifiers on this small-scale dataset regarding overall classification performance. It was also able to identify some typical morphological characteristics in H&E images to provide histopathological interpretations for pathologists. KeywordsEndometrial cancer; Hematoxylin and eosin (H&E) images; Deep learning; Class activation map (CAM); Human-machine collaboration. Reshape input target_shape = [-1,256] Reshape input target_shape = [-1,256] MaxPooling1D input strides = [2] pool_siz...

By applying the time-delay control theory to a TCP/RED dynamic model, this note establishes some explicit conditions under which the TCP/RED system is stable in terms of the average queue length. Then, the stability region is discussed. Finally, the results are illustrated by using 2 simulations, which demonstrates that it is able to choose an appropriate control parameter max of RED based on the stability conditions derived in this note, to achieve satisfactory network performance. It is found, by comparison, that this improved performance is better than that of three other typical active queue management (AQM) schemes-the random exponential marking (REM), proportional-integral (PI) controller, and adaptive virtual queue (AVQ) schemes.Index Terms-Active queue management (AQM), random early detection (RED), stability, time-delay control.

No abstract

No abstract

Adaptive Monte Carlo localization (AMCL) algorithm has a limited pose accuracy because of the nonconvexity of the laser sensor model, the complex and unstructured features of the working environment, the randomness of particle sampling, and the final pose selection problem. In this paper, an improved AMCL algorithm is proposed, aiming to build a laser radar-based robot localization system in a complex and unstructured environment, with a LIDAR point cloud scan-matching process after the particle score calculating process. The weighted mean pose of AMCL particle swarm is used as the initial pose of the scan matching process. The LIDAR point cloud is matched with the probability grid map from coarse to fine using the Gaussian-Newton method, which results in more accurate poses. Moreover, the scan-matching pose is added into the particle swarm as a high-weight particle. So the particle swarm after resampling will be more concentrated in the correct position. The particle filter and the scan-matching process form a closed loop, thus enhancing the localization accuracy of mobile robots. The experiment results demonstrate that the proposed improved AMCL algorithm is superior to the traditional AMCL algorithm in the complex and unstructured environment, by exploiting the high-accuracy characteristic of scan matching while inheriting the stability of AMCL.