Mortality from lung cancer has ranked high among cancers for many years. Early detection of lung cancer is critical for disease prevention, cure, and mortality rate reduction. Many existing detection methods on lung nodules can achieve high sensitivity but meanwhile introduce an excessive number of false-positive proposals, which is clinically unpractical. In this paper, we propose the multi-head detection and spatial attention network, shortly MHSnet, to address this crucial false-positive issue. Specifically, we first introduce multi-head detectors and skip connections to capture multi-scale features so as to customize for the variety of nodules in sizes, shapes, and types. Then, inspired by how experienced clinicians screen CT images, we implemented a spatial attention module to enable the network to focus on different regions, which can successfully distinguish nodules from noisy tissues. Finally, we designed a lightweight but effective false-positive reduction module to cut down the number of false-positive proposals, without any constraints on the front network. Compared with the state-of-the-art models, our extensive experimental results show the superiority of this MHSnet not only in the average FROC but also in the false discovery rate (2.64% improvement for the average FROC, 6.39% decrease for the false discovery rate). The false-positive reduction module takes a further step to decrease the false discovery rate by 14.29%, indicating its very promising utility of reducing distracted proposals for the downstream tasks relied on detection results.
As XML is playing a crucial role in web services, databases, and document processing, efficient processing of XML queries has become an important issue. On the other hand, due to the increasing number of users, high throughput of XML queries is also required to execute tens of thousands of queries in a short time. Given the great success of GPGPU (General-Purpose computations on the Graphics Processors), we propose a parallel XML query model based on GPU, which mainly consists of two efficient task distribution strategies, to improve the efficiency and throughput of XML queries. We have developed a parallel simplified XPath language using Compute Unified Device Architecture (CUDA) on GPU, and evaluate our model on a recent NVIDIA GPU in comparison with its counterpart on eight-core CPU. The experiment results show that our model achieves both higher throughput and efficiency than CPU-based XML query.
At present, there are three main mobile apps marketplaces, iTunes App Store, Android Market and Windows Phone Store. With app recommendation technology, users not only discover more relevant apps, but they're also more likely to be engaged with those apps on a higher level because they are relevant to their interests in the first place. Collaborative filtering (CF) methods had been applied to recommender systems, but the CF techniques do not handle sparse dataset well, especially in the case of the cold start problem where there is no enough interaction for apps. To conquer this constraint, we propose a novel recommending model: Interoperability-Enriched Recommendation (IER) that is an interoperability-enriched collaborative filtering method for multi-marketplace app recommendation based on the global app ecosystem. Experimental results on the known marketplaces app dataset demonstrate that the proposed IER method significantly outperforms the state-of-the-art CF method and context-aware recommendations (CAR) method for app recommendation, especially in the cold start scenario.
Considering the complex uncertain database, top-kquery processing in uncertain databases is semantically and computationally different from classical top-kprocessing. Score is not the only factor we should concern. The interplay between score and membership uncertainty makes computation complex. Powerful computing capability of Graphic Processing Unit(GPU) is needed in the processing of this kind of queries if we want to acquire the results as soon as possible. Using GPU with batch mode, we present a CPUGPU cooperative computing framework to processing top-k queries in uncertain database. Two parallel GPU algorithms are designed to solve the problem specifically. Moreover, a "label-confidence" data format conversion is proposed to reduce CPU-GPU communication. We also suggest an error correction method with the heap-based algorithm to improve the accuracy and correction of the results. Experimental results show that the CPU-GPU framework provides a better performance and it is quite efficiency in handling uncertain top-k problem.
The mortality of lung cancer has ranked high among cancers for many years. Early detection of lung cancer is critical for disease prevention, cure, and mortality rate reduction. However, existing detection methods on pulmonary nodules introduce an excessive number of false positive proposals in order to achieve high sensitivity, which is not practical in clinical situations. In this paper, we propose the multi-head detection and spatial squeeze-and-attention network, MHSnet, to detect pulmonary nodules, in order to aid doctors in the early diagnosis of lung cancers. Specifically, we first introduce multi-head detectors and skip connections to customize for the variety of nodules in sizes, shapes and types and capture multi-scale features. Then, we implement a spatial attention module to enable the network to focus on different regions differently inspired by how experienced clinicians screen CT images, which results in fewer false positive proposals. Lastly, we present a lightweight but effective false positive reduction module with the Linear Regression model to cut down the number of false positive proposals, without any constraints on the front network. Extensive experimental results compared with the state-of-the-art models have shown the superiority of the MHSnet in terms of the average FROC, sensitivity and especially false discovery rate (2.98% and 2.18% improvement in terms of average FROC and sensitivity, 5.62% and 28.33% decrease in terms of false discovery rate and average candidates per scan). The false positive reduction module significantly decreases the average number of candidates generated per scan by 68.11% and the false discovery rate by 13.48%, which is promising to reduce distracted proposals for the downstream tasks based on the detection results.
In recent years, the mortality rate of respiratory diseases ranks high among the major diseases. Early detection of respiratory diseases is a key factor in reducing the mortality rate and curing diseases. In this paper, we propose the BiGRU Attention-XGBoost model to classify respiratory sounds, in order to assist doctors in the early diagnosis of respiratory diseases. Specifically, we first extract two sets of features, i.e., the time domain and spectral features to encode respiratory sounds. Then, we apply the Gradient Boosting Decision Tree algorithm to select important features for classification. Based on the temporal characteristics of respiratory sounds, we design the BiGRU Attention-XGBoost model to classify them. Finally, to enlarge the training dataset and address the problem of data imbalance, we also implement Griffifin-Lim and WORLD Vocoder, two data augmentation methods. Extensive experiments show the superiority of the proposed model compared to seven state-of the-art models in terms of classification accuracy and Fl-score.
The medical community question answering system (MCQA) which is a novel medical communicating platform is becoming popular due to the large amount of people seeking for solution of health problem. Not all medical questions would get timely answers. Similar question recommendation is a common approach to solve this problem. While the large amount of Q&A pairs in MCQA, classifying for a new problem effectively reduces the size of candidate problem set and improve the efficiency of retrieval for its similar problems. In this paper we proposed a shallow CNN classification model (SCCM) based on characters which not only narrowing the text feature vector dimensions, but also solving spelling mistakes in text. We crawls and constructs a real Chinese medical Q&A dataset and conducts experimental verification. The results of experiments show that our SCCM can automatically extract the key features in the text and improve the accuracy of classification.
In this paper, a wheeled mobile robot system consisting of a tractor and N-1 trailers is studied. Based on Kanes approach, the dynamics equations of the system is established, especially the coefficients are computed using recursive algorithm. In this way, the reconstruction of the dynamics is discussed further after the number of the trailers has been changed.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)