Search

This study aimed to evaluate the incidence and prognosis of primary cardiac lymphoma (PCL) by using the Surveillance, Epidemiology, and End Results Program (SEER) database. Patients diagnosed with PCL and the disease incidence in the SEER database from 1975 to 2016 were included. Overall survival (OS) and cause-specific survival (CSS) curves were calculated using the Kaplan-Meier method and compared by the log-rank test. Univariate and multivariable Cox proportional hazard regression analyses were used to identify associations with outcome measures. The incidence of PCL was 0.011/100 000, and a predominance of elderly and male patients was observed. A total of 144 patients were enrolled. The median age of onset was 68 (9-96) years, including 80 (55.6%) males and 64 (44.4%) females. Multivariate analysis revealed that age and chemotherapy were independent prognostic factors for OS (both P < .05). Ann Arbor stage and chemotherapy were independent prognostic factors for CSS (both P < .05). In terms of treatment modality, chemotherapy combined with surgery was an independent protective factor for OS and CSS (both P < .05). For patients with primary cardiac diffuse large B-cell lymphoma (cardiac DLBCL), multivariate analysis also showed that age, Ann Arbor stage, and chemotherapy were all independent prognostic factors for OS and CSS (all P < .05). Chemotherapy combined with surgery was associated with a significant benefit in terms of OS and CSS (both P < .05). Our study confirmed that older age and advanced Ann Arbor stage were independent risk factors for PCL, and treatment with chemotherapy or cooperation with surgery resulted in better long-term survival.

Pedestrian detection has achieved great improvements with the help of Convolutional Neural Networks (CNNs). CNN can learn high-level features from input images, but the insufficient spatial resolution of CNN feature channels (feature maps) may cause a loss of information, which is harmful especially to small instances. In this paper, we propose a new pedestrian detection framework, which extends the successful RPN+BF framework to combine handcrafted features and CNN features. RoI-pooling is used to extract features from both handcrafted channels (e.g. HOG+LUV, CheckerBoards or RotatedFilters) and CNN channels. Since handcrafted channels always have higher spatial resolution than CNN channels, we apply RoI-pooling with larger output resolution to handcrafted channels to keep more detailed information. Our ablation experiments show that the developed handcrafted features can reach better detection accuracy than the CNN features extracted from the VGG-16 net, and a performance gain can be achieved by combining them. Experimental results on Caltech pedestrian dataset with the original annotations and the improved annotations demonstrate the effectiveness of the proposed approach. When using a more advanced RPN in our framework, our approach can be further improved and get competitive results on both benchmarks., Comment: 9 pages, 4 figures, Submitted to Neurocomputing, The 5th line of table 3 was accidentally mistaken. The data have been corrected and the related descriptions in section 4.4 have also be revised accordingly. Typos corrected, references corrected

Pedestrian detection is a fundamental component in many real-world applications such as automatic driving, intelligent surveillance, person re-identification and robotics. Therefore, it has attracted massive attention in the last decades. However, pedestrians in an images always exhibit different scales, which constitutes a significant mode of intra-class variability and affect the performance of pedestrian detection algorithm. To address this problem, we apply FPN (Feature Pyramid Network)for pedestrian detection. FPN exploits the inherent multi-scale structure of a deep convolutional network to construct a feature pyramid that has rich semantics at all levels and facilitates the detection of objects at different scales. To leverage the information from different levels of the feature pyramid, we extend the FPN-based pedestrian detection by fusing the feature of each level with adaptive feature pooling. Furthermore, we also integrate a Squeeze and Excitation module to the ROI pooled features from each level before the feature fusion. The experiment result on Caltech dataset shows that our approach outperforms the basic FPN-based pedestrian detection and robust towards to various scale of pedestrian.