Achieving the pathological type diagnosis of pulmonary nodules on chest CT is a critical step in the early detection of lung cancer and treatment of patients. Based on a small and unbalanced self-constructed dataset, we achieved intelligent diagnosis of five pathological types including adenocarcinoma, squamous cell carcinoma, small cell carcinoma, inflammatory and other benign diseases for the first time. In order to reduce the dependence of deep convolutional neural network (DCNN) on a large amount of training data, a reverse adversarial classification network (RACN) was proposed based on semi-supervised learning, which consists of a reverse generative adversarial network (RGAN) for unsupervised regression and a supervised classification network (CN). In RGAN, five specific normal distributions P with different means and variances were assigned to represent the five pathological types, and then a special regression task was designed by mapping pulmonary nodules to the random sampling Z of P. The input of generator in RGAN is set to 3D nodule volume data, the inputs of discriminator are set to Z and the output of generator. The regression task enables RGAN to extract specific features, which will be deeply integrate into CN to improve the classification performance. Experiments showed that the average sensitivity of RACN in detecting malignant nodules was 0.6525, where the sensitivity of adenocarcinoma, small cell carcinoma and squamous cell carcinoma was 0.8426, 0.5604 and 0.5543. Besides, the RACN can achieve 93.21% accuracy for diagnosing malignant nodules on the public LIDC-IDRI dataset, obtaining the state-of-the-art results.

To accurately diagnose pneumonia patients on a limited annotated chest X-ray image dataset, a prior knowledge-based active attention network (PKA2-Net1) was constructed. The PKA2-Net uses improved ResNet as the backbone network and consists of residual blocks, novel subject enhancement and background suppression (SEBS) blocks and candidate template generators, where template generators are designed to generate candidate templates for characterizing the importance of different spatial locations in feature maps. The core of PKA2-Net is SEBS block, which is proposed based on the prior knowledge that highlighting distinctive features and suppressing irrelevant features can improve the recognition effect. The purpose of SEBS block is to generate active attention features without any high-level features and enhance the ability of the model to localize lung lesions. In SEBS block, first, a series of candidate templates T with different spatial energy distributions are generated and the controllability of the energy distribution in T enables active attention features to maintain the continuity and integrity of the feature space distributions. Second, Top-ntemplates are selected from T according to certain learning rules, which are then operated by a convolution layer for generating supervision information that can guide the inputs of SEBS block to form active attention features. We evaluated the PKA2-Net on the binary classification problem of identifying pneumonia and healthy controls on a dataset containing 5856 chest X-ray images (ChestXRay2017), the results showed that our method can achieve 97.63% accuracy and 0.9872 sensitivity.

In medical image analysis, in order to reduce the impact of unbalanced data sets on data-driven deep learning models, according to the characteristic that the area under the Precision-Recall curve (AUCPR) is sensitive to each category of samples, a novel Harmony loss function with fast convergence speed and high stability was constructed. Since AUCPRneeds to be calculated in discrete domain, in order to ensure the continuous differentiability and gradient existence of the Harmony loss, first, the Logistic function was used to approximate the Logical function in AUCPR. Then, to improve the optimization speed of the Harmony loss during model training, a method of manually setting a certain number of classification thresholds was proposed to further approximate the calculation of AUCPR. After the above two approximate calculation processes, the Harmony loss with stable gradient and high computational efficiency was designed. In the optimization process of the model, since Harmony loss can reconcile recall and precision of each category under different classification thresholds, thereby, it can not only improve the model's ability to recognize categories with less samples, but also maintain the stability of the training curve. To comprehensively evaluate the effects of Harmony loss function, we performed experiments on image 3D reconstruction, 2D segmentation, and unbalanced classification tasks. Experimental results showed that the Harmony loss achieved the state-of-the-art results on four unbalanced data sets. Moreover, the Harmony loss can be easily combined with existing loss functions, and is suitable for most common deep learning models.

Respiratory motion estimation is an important part in image-guided radiation therapy and clinical diagnosis. However, most of the respiratory motion estimation methods rely on indirect measurements of external breathing indicators, which will not only introduce great estimation errors, but also bring invasive injury for patients. In this paper, we propose a method of lung respiratory motion estimation based on fast Kalman filtering and 4D CT image registration (LRME-4DCT). In order to perform dynamic motion estimation for continuous phases, a motion estimation model is constructed by combining two kinds of GPU-accelerated 4D CT image registration methods with fast Kalman filtering method. To address the high computational requirements of 4D CT image sequences, a multi-level processing strategy is adopted in the 4D CT image registration methods, and respiratory motion states are predicted from three independent directions. In the DIR-lab dataset and POPI dataset with 4D CT images, the average target registration error (TRE) of the LRME-4DCT method can reach 0.91 mm and 0.85 mm respectively. Compared with traditional estimation methods based on pair-wise image registration, the proposed LRME-4DCT method can estimate the physiological respiratory motion more accurately and quickly. Our proposed LRME-4DCT method fully meets the practical clinical requirements for rapid dynamic estimation of lung respiratory motion.

To solve the problem that traditional image registration methods based on continuous optimization for large motion lung 4D CT image sequences are easy to fall into local optimal solutions and lead to serious misregistration, a novel image registration method based on high-order Markov Random Field (MRF) is proposed. By analyzing the effect of the deformation field constraint of the potential functions with different order cliques in MRF model, energy functions with high-order cliques form are designed separately for 2D and 3D images to preserve topology of the deformation field. In order to preserve the topology of the deformation field more effectively, it is necessary to apply a smooth term and a topology preservation term simultaneously in the energy function and use logarithmic function to impose a penalty on the Jacobian matrix with high-order cliques in the topology preservation term. For the complexity of the designed energy function with high-order cliques form, Markov Chain Monte Carlo (MCMC) algorithm is used to solve the optimization problem of the designed energy function. To address the high computational requirements in lung 4D CT image registration, a multi-level processing strategy is adopted to reduce the space complexity of the proposed registration method and promotes the computational efficiency. In the DIR-lab dataset with 4D CT images and the COPD (Chronic Obstructive Pulmonary Disease) dataset with 3D CT images, the average target registration error (TRE) of our proposed method can reach 0.95 mm respectively.

The time synchronization of network will encounter a severe challenge due to the proliferation of scale in wireless sensor networks. In order to improve the precision of time synchronization in network, a time synchronization protocol based on frequency skew bid and dynamic topology (FBDT) is designed in the paper as well as realized on Embedded Linux System. To be able to effectively inhibit synchronization error accumulation in multi-hop, a clustering mechanism in network region is adopted to reduce synchronization hops. In the protocol, every node selects the parent node which has a more stable clock frequency and fewer synchronization hops to restrain accumulation of synchronization error, meanwhile, combining fast and slow synchronization strategy is utilized to meet requirement on two performance indicators of synchronization accuracy and convergence time. Availing itself of hops information to select the parent node dynamically, the protocol can fit well in the network of dynamic topology, with good resistance to destructive and scalability. A method similar with TPSN protocol is used to eliminate transmission delay during synchronization between two nodes. Testing results on embedded Linux development board indicate that the FBDT synchronization protocol can reach the precision of μs, which can satisfy the requirements for practical applications.

In order to solve the problem the homogeneous coefficient of the classic active demons algorithm can not take into account large deformation and small deformation at the same time, this paper presents a non-rigid registration algorithm based on active demons algorithm. The proposed algorithm introduces a new parameter called balance coefficient to the active demons algorithm, which will adjust the driving force combined with homogeneous coefficient. Not only the large deformation and the small deformation can be taken into account at the same time, but also the mutual restraint problem of the convergence speed and the registration accuracy can be eased in a certain extent. In order to further improve the registration accuracy and the convergence speed, and avoid falling into local extreme value, a coarse-to-fine multi-resolution strategy is introduced into the registration process. Experiments on checkboard test images, natural images and medical images demonstrate that the proposed method is faster and more accurate, and the registration accuracy is close to the latest TV-L1 optical flow image registration algorithm, which solves the problems of the active demons algorithm.

Node flip ambiguity is a key problem that needs to be addressed for range-based node localization in wireless sensor networks. In this paper we have implemented robustness analysis for three-dimensional node localization in wireless sensor networks. A robustness criterion to detect flip ambiguity for range-based three-dimensional nodes localization is proposed. We have proposed that flip ambiguity detection for three-dimensional node localization is equal to whether there is a plane intersecting with all range error spheres of the reference nodes of the unknown node, which is called the existence of intersecting plane (EIP) problem. To solve EIP problem, we further have proposed two solving algorithms: common tangent plane algorithm (CTP) and orthogonal projection algorithm (OP). The simulation experiments demonstrate that CTP has good detection results, but its computational complexity is too high; however, OP has almost the same detection results as CTP and has lower computational complexity.

In this paper, we propose a node localization algorithm based on the received signal strength (RSS) measurements and the Leader Intelligent Selection (LIS) optimization algorithm in Wireless Sensor Networks (WSN). The LIS optimization algorithm is proposed based on the idea of biological heuristic. By designing a simple animal group leader se lection mode, a leader candidates' group is searched by the leader searcher, and an optimal individual is selected from the group as the leader which is the global optimal solution of the optimization problem by evaluating each leader candidate's ability. In order to accele rate the leader's campaign and the evolutionary rate in the later period of LIS, the simple Minimum Mean Square Error (MMSE) algorithm or the centroid algorithm is adopted to obtain an initial coordinate as the initial leader of LIS algorithm using the information of the anchor node coordinates and the ranging findings. By considering fully the distance factor, an improved objective function is defined, so the node localization problem in WSN could be transformed into a nonlinear unconstrained optimization problem. The proposed LIS algorithm is used to solve this problem, and the obtained solution is the estimated value of the WSN node's coordinates. Compared with the Artificial Bee Colony (ABC) algorithm, the Particle Swarm Optimization (PSO) algorithm and the Genetic Algorithm (GA), the proposed LIS algorithm is better than the others in accuracy and calculation complexity.

In order to reduce the error accumulation of time synchronization for large-scale and multi-hop wireless sensor networks, a time synchronization protocol based on dynamic route list (DRL-TSP) is proposed. The protocol uses an available dynamic route list which exists in each node, and allows nodes to choose an optimal synchronization route adaptively to reduce error accumulation. An identification named as Time to Available (TTA) is utilized to deal with the synchronization failure and ensure the reliability of the synchronization process. The proposed protocol is applied in the level-based wireless sensor network and simulated by NS2. Experimental results demonstrate that the proposed protocol can inhibit the synchronization error accumulation effectively and has better performance than the current protocol in precision and reliability.

To solve the problem of slow convergence using the Sage-Husa algorithm to estimate clock frequency on time synchronization in wireless sensor networks, a bimodal clock frequency estimation algorithm is proposed. The algorithm uses a threshold to determine which mode the clock frequency is. For the jump mode, the ratio-based method is adopted to track the clock frequency quickly. For the stable mode, the Sage-Husa method is used to improve the estimated accuracy. Finally the proposed algorithm is applied in the level-based wireless sensor network and simulated on NS2. Experimental results show that the proposed method has better performance than the current algorithms (the ratio-based method, the Least Square Algorithm, and so on) in precision and convergence.

This paper presents an energy-efficient distributed link scheduling protocol based on vertex coloring method for wireless sensor networks. In the protocol, we perform vertex coloring using maximum degree preferred scheme, and gather network information only by local message exchanges. Thus, our protocol can further reduce the maximum number of timeslots, and avoid long-distance multihop packet forwarding. To improve coloring success rate, two mechanisms are proposed in the protocol. The first is a broadcast guarantee mechanism for increasing broadcast message delivery rate; the second is a conflict processing mechanism for solving coloring conflicts. Simulation results show that the proposed distributed scheduling protocol has better performance than certain centralized scheduling protocols and a classical distributed protocol DRAND in energy efficiency, spatial reuse rate and packet loss rate.

To the problem of no overall optimal merger for one-way merger in the segmentation algorithm proposed by Wang et al., (1999), we propose a method of overall optimal search and merger. At the same time, for the problem of merging a segment which has non-value (value-segment) and a segment whose values are zeros entirely (zeros-segment) to a large segment in Wang's method, we also propose a corresponding method to solve the problem. The main techniques use the local cosine transform (LCT) algorithm for a single small segment, rather than folding processing using its original neighboring data, instead of making zero extension, and then fold the each zero-extension segment. A great deal of numerical simulations validate that this new improved technique solves several problems of the binary-based segment algorithm and Wang's segment algorithm; it not only obtains adapted effective segmentation results, but also there are not many redundancy segmentations.