EFNet: enhancing feature information for 3D object detection in LiDAR point clouds.

With the development of autonomous driving, there has been considerable attention on 3D object detection using LiDAR. Pillar-based LiDAR point cloud detection algorithms are extensively employed in the industry due to their simple structure and high real-time performance. Nevertheless, the pillar-based detection network suffers from significant loss of 3D coordinate information during the feature degradation and extraction process. In the paper, we introduce a novel framework with high performance, termed EFNet. The EFNet uses the Enhancing Pillar Feature Module (EPFM) to provide more accurate representations of features from two directions: pillar internal space and pillar external space. Additionally, the Head Up Module (HUM) is utilized in the detection head to integrate multi-scale information and enhance the network's information perception ability. The EFNet achieves impressive results on the nuScenes datasets, namely, 53.3% NDS and 42.4% mAP. Compared to the baseline PointPillars, EFNet improves 8% NDS and 11.9% mAP. The results demonstrate that the proposed framework can effectively improve the network's accuracy while ensuring deployability.

TS-LCD: Two-Stage Loop-Closure Detection Based on Heterogeneous Data Fusion.

Loop-closure detection plays a pivotal role in simultaneous localization and mapping (SLAM). It serves to minimize cumulative errors and ensure the overall consistency of the generated map. This paper introduces a multi-sensor fusion-based loop-closure detection scheme (TS-LCD) to address the challenges of low robustness and inaccurate loop-closure detection encountered in single-sensor systems under varying lighting conditions and structurally similar environments. Our method comprises two innovative components: a timestamp synchronization method based on data processing and interpolation, and a two-order loop-closure detection scheme based on the fusion validation of visual and laser loops. Experimental results on the publicly available KITTI dataset reveal that the proposed method outperforms baseline algorithms, achieving a significant average reduction of 2.76% in the trajectory error (TE) and a notable decrease of 1.381 m per 100 m in the relative error (RE). Furthermore, it boosts loop-closure detection efficiency by an average of 15.5%, thereby effectively enhancing the positioning accuracy of odometry.

Compressed Gaussian Estimation under Low Precision Numerical Representation.

This paper introduces a novel method for computationally efficient Gaussian estimation of high-dimensional problems such as Simultaneous Localization and Mapping (SLAM) processes and for treating certain Stochastic Partial Differential Equations (SPDEs). The authors have presented the Generalized Compressed Kalman Filter (GCKF) framework to reduce the computational complexity of the filters by partitioning the state vector into local and global and compressing the global state updates. The compressed state update, however, still suffers from high computational costs, making it challenging to implement on embedded processors. We propose a low-precision numerical representation for the global filter, such as 16-bit integer or 32-bit single-precision formats for the global covariance matrix, instead of the expensive double-precision, floating-point representation (64 bits). This truncation can inevitably cause filter instability since the truncated covariance matrix becomes overoptimistic or even turns to be an invalid covariance matrix. We introduce a Minimal Covariance Inflation (MCI) method to make the filter consistent while minimizing the truncation errors. Simulation-based experiments results show significant improvement of the proposed method with a reduction in the processing time with minimal loss of accuracy.

Efficient Clustering for Continuous Occupancy Mapping Using a Mixture of Gaussian Processes.

This paper proposes a novel method for occupancy map building using a mixture of Gaussian processes. Gaussian processes have proven to be highly flexible and accurate for a robotic occupancy mapping problem, yet the high computational complexity has been a critical barrier for large-scale applications. We consider clustering the data into small, manageable subsets and applying a mixture of Gaussian processes. One of the problems in clustering is that the number of groups is not known a priori, thus requiring inputs from experts. We propose two efficient clustering methods utilizing (1) a Dirichlet process and (2) geometrical information in the context of occupancy mapping. We will show that the Dirichlet process-based clustering can significantly speed up the training step of the Gaussian process and if geometrical features, such as line features, are available, they can further improve the clustering accuracy. We will provide simulation results, analyze the performance and demonstrate the benefits of the proposed methods.

Depth-Camera-Aided Inertial Navigation Utilizing Directional Constraints.

This paper presents a practical yet effective solution for integrating an RGB-D camera and an inertial sensor to handle the depth dropouts that frequently happen in outdoor environments, due to the short detection range and sunlight interference. In depth drop conditions, only the partial 5-degrees-of-freedom pose information (attitude and position with an unknown scale) is available from the RGB-D sensor. To enable continuous fusion with the inertial solutions, the scale ambiguous position is cast into a directional constraint of the vehicle motion, which is, in essence, an epipolar constraint in multi-view geometry. Unlike other visual navigation approaches, this can effectively reduce the drift in the inertial solutions without delay or under small parallax motion. If a depth image is available, a window-based feature map is maintained to compute the RGB-D odometry, which is then fused with inertial outputs in an extended Kalman filter framework. Flight results from the indoor and outdoor environments, as well as public datasets, demonstrate the improved navigation performance of the proposed approach.

Uncertainty Estimation of Dense Optical Flow for Robust Visual Navigation.

This paper presents a novel dense optical-flow algorithm to solve the monocular simultaneous localisation and mapping (SLAM) problem for ground or aerial robots. Dense optical flow can effectively provide the ego-motion of the vehicle while enabling collision avoidance with the potential obstacles. Existing research has not fully utilised the uncertainty of the optical flow-at most, an isotropic Gaussian density model has been used. We estimate the full uncertainty of the optical flow and propose a new eight-point algorithm based on the statistical Mahalanobis distance. Combined with the pose-graph optimisation, the proposed method demonstrates enhanced robustness and accuracy for the public autonomous car dataset (KITTI) and aerial monocular dataset.

Ingested d-Aspartate Facilitates the Functional Connectivity and Modifies Dendritic Spine Morphology in Rat Hippocampus.

d-Aspartate (d-Asp), the stereoisomer of l-aspartate, has a role in memory function in rodents. However, the mechanism of the effect of d-Asp has not been fully understood. In this study, we hypothesized that ingested d-Asp directly reaches the hippocampal tissues via the blood circulation and modifies the functional connectivity between hippocampus and other regions through spinogenesis in hippocampal CA1 neurons. The spinogenesis induced by the application of d-Asp was investigated using rat acute hippocampal slices. The density of CA1 spines was increased following 21 and 100 µM d-Asp application. The nongenomic spine increase pathway involved LIM kinase. In parallel to the acute slice study, brain activation was investigated in awake rats using functional MRI following the intragastric administration of 5 mM d-Asp. Furthermore, the concentration of d-Asp in the blood serum and hippocampus was significantly increased 15 min after intragastric administration of d-Asp. A functional connectivity by awake rat fMRI demonstrated increased slow-frequency synchronization in the hippocampus and other regions, including the somatosensory cortex, striatum, and the nucleus accumbens, 10-20 min after the start of d-Asp administration. These results suggest that ingested d-Asp reaches the brain through the blood circulation and modulates hippocampal neural networks through the modulation of spines.

Interleukin-8 in the Tumor Immune Niche: Lessons from Comparative Oncology.

Interleukin (IL)-8 is a chemokine that is essential for inflammation and angiogenesis. IL-8 expression is elevated in tumor cell lines and tissues, as well as in peripheral blood obtained from cancer patients. Primary works have attempted to determine the biological effect of IL-8 on tumor cells, including cell proliferation, survival, and migration. More recently, IL-8 has acquired considerable attention as an immune modulator in the context of certain tumor microenvironments (TME); specifically, it can support a niche that favors tumor progression and metastasis. Tumor-derived IL-8 stimulates inflammation by interacting with the microenvironmental constituents, including fibroblasts, endothelial cells, and immune cells. However, the tumor immune system is complex, and mechanisms that construct the immune phenotype remain incompletely characterized. Herein, we will (1) address a potential role of IL-8 in regulating gene expression to establish immune landscape in tumor. Then, we will (2) review IL-8 signaling in the maintenance of stem cells and regulation of hematopoietic progenitors. Finally, (3) IL-8 functions will be discussed in naturally occurring animal cancers that offer a clinically realistic model for translational research. This chapter will provide a new insight into the tumor immune niche and help us develop immunotherapies for cancers.

Somatic mutational profiles of stage II and III gastric cancer according to tumor microenvironment immune type.

We aimed to determine somatic mutational profiles of stage II/III gastric cancers (GCs) according to their tumor microenvironment immune types (TMITs), which classify cancer based on co-assessment of PD-L1 expression and CD8+ tumor infiltrating lymphocytes. Eighty patients with stage II/III GC were classified as follows: TMIT I (PD-L1+ /CD8High ), TMIT II (PD-L1- /CD8Low ), TMIT III (PD-L1+ /CD8Low ), and TMIT IV (PD-L1- /CD8High ). Deep targeted sequencing using a panel of 170 cancer-related genes was performed on an Illumina HiSeq-2500 system. Most frequently mutated genes included GNAQ (41.3%), TP53 (38.8%), CREBBP (35.0%), and MAP3K1 (35.0%). PIK3CA mutations were observed more frequently in TMIT I (45.8%) and III (66.7%), than in II (12.0%) and IV (8.0%). Other genes with enriched mutations within TMIT I included ATM (33.3%), BRCA2 (33.3%), MAP3K4 (29.2%), and FLT4 (25.0%). FGFR3, MAP3K1, and RUNX1 mutations were more frequently found in TMIT II. TMIT III had a unique somatic mutation profile harboring enriched mutations of histone modifiers including CREBBP and KMT2A, and we found FGFR2 amplification exclusively within TMIT IV. Fuzzy clustering analysis based on somatic mutation frequencies identified a hypermutated group (cluster 1) and a hypomutated group (cluster 2). Cluster 1 had significant associations with TMIT I, EBV+ GCs, and MSI-H GCs (P = .023, .014, and .004), and had better overall survival (P = .057) than Cluster 2. TMIT I, EBV+ , and MSI-H GCs were estimated to have greater tumor mutational burden (P = .023, .003, and .015). By analyzing somatic mutation profiles according to TMIT classification, we identified TMIT-specific genetic alterations that provide clues for biological linkage between GC genetics and microenvironment.

Genome-wide association study identifies shared risk loci common to two malignancies in golden retrievers.

Dogs, with their breed-determined limited genetic background, are great models of human disease including cancer. Canine B-cell lymphoma and hemangiosarcoma are both malignancies of the hematologic system that are clinically and histologically similar to human B-cell non-Hodgkin lymphoma and angiosarcoma, respectively. Golden retrievers in the US show significantly elevated lifetime risk for both B-cell lymphoma (6%) and hemangiosarcoma (20%). We conducted genome-wide association studies for hemangiosarcoma and B-cell lymphoma, identifying two shared predisposing loci. The two associated loci are located on chromosome 5, and together contribute ~20% of the risk of developing these cancers. Genome-wide p-values for the top SNP of each locus are 4.6×10-7 and 2.7×10-6, respectively. Whole genome resequencing of nine cases and controls followed by genotyping and detailed analysis identified three shared and one B-cell lymphoma specific risk haplotypes within the two loci, but no coding changes were associated with the risk haplotypes. Gene expression analysis of B-cell lymphoma tumors revealed that carrying the risk haplotypes at the first locus is associated with down-regulation of several nearby genes including the proximal gene TRPC6, a transient receptor Ca2+-channel involved in T-cell activation, among other functions. The shared risk haplotype in the second locus overlaps the vesicle transport and release gene STX8. Carrying the shared risk haplotype is associated with gene expression changes of 100 genes enriched for pathways involved in immune cell activation. Thus, the predisposing germ-line mutations in B-cell lymphoma and hemangiosarcoma appear to be regulatory, and affect pathways involved in T-cell mediated immune response in the tumor. This suggests that the interaction between the immune system and malignant cells plays a common role in the tumorigenesis of these relatively different cancers.

Sensor-Based Optimization Model for Air Quality Improvement in Home IoT.

We introduce current home Internet of Things (IoT) technology and present research on its various forms and applications in real life. In addition, we describe IoT marketing strategies as well as specific modeling techniques for improving air quality, a key home IoT service. To this end, we summarize the latest research on sensor-based home IoT, studies on indoor air quality, and technical studies on random data generation. In addition, we develop an air quality improvement model that can be readily applied to the market by acquiring initial analytical data and building infrastructures using spectrum/density analysis and the natural cubic spline method. Accordingly, we generate related data based on user behavioral values. We integrate the logic into the existing home IoT system to enable users to easily access the system through the Web or mobile applications. We expect that the present introduction of a practical marketing application method will contribute to enhancing the expansion of the home IoT market.

Sensor-Based Real-Time Detection in Vulcanization Control Using Machine Learning and Pattern Clustering.

Recent paradigm shifts in manufacturing have resulted from the need for a smart manufacturing environment. In this study, we developed a model to detect anomalous signs in advance and embedded it in an existing programmable logic controller system. For this, we investigated the innovation process for smart manufacturing in the domain of synthetic rubber and its vulcanization process, as well as a real-time sensing technology. The results indicate that only analysis of the pattern of input variables can lead to significant results without the generation of target variables through manual testing of chemical properties. We have also made a practical contribution to the realization of a smart manufacturing environment by building cloud-based infrastructure and models for the pre-detection of defects.

Identification of three molecular and functional subtypes in canine hemangiosarcoma through gene expression profiling and progenitor cell characterization.

Canine hemangiosarcomas have been ascribed to an endothelial origin based on histologic appearance; however, recent findings suggest that these tumors may arise instead from hematopoietic progenitor cells. To clarify this ontogenetic dilemma, we used genome-wide expression profiling of primary hemangiosarcomas and identified three distinct tumor subtypes associated with angiogenesis (group 1), inflammation (group 2), and adipogenesis (group 3). Based on these findings, we hypothesized that a common progenitor may differentiate into the three tumor subtypes observed in our gene profiling experiment. To investigate this possibility, we cultured hemangiosarcoma cell lines under normal and sphere-forming culture conditions to enrich for tumor cell progenitors. Cells from sphere-forming cultures displayed a robust self-renewal capacity and exhibited genotypic, phenotypic, and functional properties consistent with each of the three molecular subtypes seen in primary tumors, including expression of endothelial progenitor cell (CD133 and CD34) and endothelial cell (CD105, CD146, and αvß3 integrin) markers, expression of early hematopoietic (CD133, CD117, and CD34) and myeloid (CD115 and CD14) differentiation markers in parallel with increased phagocytic capacity, and acquisition of adipogenic potential. Collectively, these results suggest that canine hemangiosarcomas arise from multipotent progenitors that differentiate into distinct subtypes. Improved understanding of the mechanisms that determine the molecular and phenotypic differentiation of tumor cells in vivo could change paradigms regarding the origin and progression of endothelial sarcomas.

Interleukin-8 promotes canine hemangiosarcoma growth by regulating the tumor microenvironment.

Interleukin-8 (IL-8) gene expression is highly up-regulated in canine hemangiosarcoma (HSA); however, its role in the pathogenesis of this disease is unknown. We investigated the expression of IL-8 in canine HSA tissues and cell lines, as well and the effects of IL-8 on canine HSA in vitro, and in vivo using a mouse xenograft model for the latter. Constitutive expression of IL-8 mRNA, IL-8 protein, and IL-8 receptor were variable among different tumor samples and cell lines, but they showed stable steady states in each cell line. Upon the addition of IL-8, HSA cells showed transient intracellular calcium fluxes, suggesting that their IL-8 receptors are functional and that IL-8 binding activates relevant signaling pathways. Yet, neither addition of exogenous IL-8 nor blockade of endogenous IL-8 by neutralizing anti-IL-8 antibody (α-IL-8 Ab) affected HSA cell proliferation or survival in vitro. To assess potential effects of IL-8 in other tumor constituents, we stratified HSA cell lines and whole tumor samples into "IL-8 high" and "IL-8 low" groups. Genome-wide gene expression profiling showed that samples in the "IL-8 high" tumor group were enriched for genes associated with a "reactive microenvironment," including activation of coagulation, inflammation, and fibrosis networks. Based on these findings, we hypothesized that the effects of IL-8 on these tumors were mostly indirect, regulating interactions with the microenvironment. This hypothesis was supported by in vivo xenograft experiments where survival and engraftment of tumor cells was inhibited by administration of neutralizing α-IL-8 Ab. Together, our results suggest that IL-8 contributes to establishing a permissive microenvironment during the early stages of tumorigenesis in HSA.

Primary central nervous system B-cell lymphoma in a young dog.

This report describes a primary central nervous system B-cell lymphoma in a 3-year-old intact female Maltese dog. Canine primary central nervous system lymphomas constitute about 4% of all intracranial primary neoplasms, but comprehensive histopathologic classifications have rarely been carried out. This is the first report of this disease in a young adult dog.

Prediction of conversion to dementia using interpretable machine learning in patients with amnestic mild cognitive impairment.

Purpose: Amnestic mild cognitive impairment (aMCI) is a transitional state between normal aging and Alzheimer's disease (AD). However, not all aMCI patients are observed to convert to AD dementia. Therefore, developing a predictive algorithm for the conversion of aMCI to AD dementia is important. Parametric methods, such as logistic regression, have been developed; however, it is difficult to reflect complex patterns, such as non-linear relationships and interactions between variables. Therefore, this study aimed to improve the predictive power of aMCI patients' conversion to dementia by using an interpretable machine learning (IML) algorithm and to identify the factors that increase the risk of individual conversion to dementia in each patient. Methods: We prospectively recruited 705 patients with aMCI who had been followed-up for at least 3 years after undergoing baseline neuropsychological tests at the Samsung Medical Center between 2007 and 2019. We used neuropsychological tests and apolipoprotein E (APOE) genotype data to develop a predictive algorithm. The model-building and validation datasets were composed of data of 565 and 140 patients, respectively. For global interpretation, four algorithms (logistic regression, random forest, support vector machine, and extreme gradient boosting) were compared. For local interpretation, individual conditional expectations (ICE) and SHapley Additive exPlanations (SHAP) were used to analyze individual patients. Results: Among the four algorithms, the extreme gradient boost model showed the best performance, with an area under the receiver operating characteristic curve of 0.852 and an accuracy of 0.807. Variables, such as age, education, the scores of visuospatial and memory domains, the sum of boxes of the Clinical Dementia Rating scale, Mini-Mental State Examination, and APOE genotype were important features for creating the algorithm. Through ICE and SHAP analyses, it was also possible to interpret which variables acted as strong factors for each patient. Conclusion: We were able to propose a predictive algorithm for each aMCI individual's conversion to dementia using the IML technique. This algorithm is expected to be useful in clinical practice and the research field, as it can suggest conversion with high accuracy and identify the degree of influence of risk factors for each patient.

Canine classical seminoma: a specific malignant type with human classifications is highly correlated with tumor angiogenesis.

BACKGROUND: Human seminoma is classified as classical seminoma (SE) and spermatocytic seminoma (SS). Human SE is known to be more malignant and metastasizing more frequently than SS. Tumor angiogenesis is highly related with tumor progression and metastasis, with microvessel density (MVD) being an important parameter of metastatic potential. Canine seminoma is not yet well-established as SE or SS type including correlation with angiogenesis. We classified canine SE and SS, and then compared them to tumor associated vessels. METHODS: Twenty-three cases of canine seminomas (2 intratubular, 9 diffuse, and 12 intratubular/diffuse seminomas showing both intratubular and diffuse patterns) were classified as SE or SS by immunohistochemistry (IHC) using monoclonal antibody against PLAP and by PAS stain. The histopathological data were then compared to see if there was a correlation with SE or SS. Angiogenesis of seminomas were evaluated by immunohistochemical assay using polyclonal antibody against Von Willebrand factor (vWF) and by calculating the means of MVD, vessels area and perimeters using computerized image analysis. Statistical Package for Social Sciences (SPSS) program was used for various statistical analyses. RESULTS: The numbers of PLAP+/PAS+ canine SEs were 8/23 (34.8%) and PLAP-/PAS- SSs were 15/23 (61.2%). All SE cases (8/8, 100%) were intratubular/diffuse types. SS types included 2 intratubular (2/15, 13.3%), 9 diffuse (9/15, 60%), and 4 intratubular/diffuse (4/15, 26.7%) types. MVD and vascular parameters in SEs were significantly higher than in SSs, showing the highest value in the intratubular/diffuse type. Seminomas observed with neoplastic cells invasion of vessels presented higher perimeter and area values than seminomas without conformed neoplastic cells invasion. CONCLUSION: In this study, we demonstrated a positive relationship between canine SE and tumor angiogenesis. Furthermore, we also showed that a tumor cells invasion of vessels were a correlated vascular parameter. Although metastasis of canine seminomas has rarely been reported, our results support that canine SE could have high metastatic potential similar to the human counterpart. Further studies are required to clarify the relationship between canine SE and clinical data with metastatic factors.

Comparative immunohistochemical characterization of canine seminomas and Sertoli cell tumors.

Primary testicular tumors are the most common causes of cancer in male dogs. Overall, the majority of canine patients should be cured by testicular surgery. However, tumor markers are not well-known in veterinary medicine. We sought to determine using immunohistochemistry whether the combined human testicular tumor markers (placental alkaline phosphatase, OCT3/4, CD30, alpha-fetoprotein, inhibin-alpha, vimentin, c-KIT, and desmin) are expressed in canine seminomas and Sertoli cell tumors (SCTs). We examined 35 canine testicular tumors, 20 seminomas and 15 SCTs. c-KIT was expressed markedly in canine seminomas. Both inhibin-alpha and vimentin were expressed significantly in canine SCTs. The results of this study demonstrate differences and similarities between tumor marker expression of testicular tumors in dogs and humans. All the main markers in current routine use are discussed as well as potential useful markers for benign and malignant tumors, and tumor progression.

Comparative Genomics Reveals Shared Mutational Landscape in Canine Hemangiosarcoma and Human Angiosarcoma.

Angiosarcoma is a highly aggressive cancer of blood vessel-forming cells with few effective treatment options and high patient mortality. It is both rare and heterogenous, making large, well-powered genomic studies nearly impossible. Dogs commonly suffer from a similar cancer, called hemangiosarcoma, with breeds like the golden retriever carrying heritable genetic factors that put them at high risk. If the clinical similarity of canine hemangiosarcoma and human angiosarcoma reflects shared genomic etiology, dogs could be a critically needed model for advancing angiosarcoma research. We assessed the genomic landscape of canine hemangiosarcoma via whole-exome sequencing (47 golden retriever hemangiosarcomas) and RNA sequencing (74 hemangiosarcomas from multiple breeds). Somatic coding mutations occurred most frequently in the tumor suppressor TP53 (59.6% of cases) as well as two genes in the PI3K pathway: the oncogene PIK3CA (29.8%) and its regulatory subunit PIK3R1 (8.5%). The predominant mutational signature was the age-associated deamination of cytosine to thymine. As reported in human angiosarcoma, CDKN2A/B was recurrently deleted and VEGFA, KDR, and KIT recurrently gained. We compared the canine data to human data recently released by The Angiosarcoma Project, and found many of the same genes and pathways significantly enriched for somatic mutations, particularly in breast and visceral angiosarcomas. Canine hemangiosarcoma closely models the genomic landscape of human angiosarcoma of the breast and viscera, and is a powerful tool for investigating the pathogenesis of this devastating disease. IMPLICATIONS: We characterize the genomic landscape of canine hemangiosarcoma and demonstrate its similarity to human angiosarcoma.

Effects of T lymphocytes, interleukin-1, and interleukin-6 on renal fibrosis in canine end-stage renal disease.

Canine end-stage renal disease (ESRD) is defined as the almost complete failure of renal function or irreversible destruction and is characterized by extensive glomerular sclerosis, tubular atrophy, interstitial inflammation, and fibrosis. Renal fibrosis is a common pathway leading to kidney failure. Infiltrating immunocytes in the end-stage kidney and several related factors are involved in renal fibrogenesis. A total of 18 renal tissue samples were obtained from canine patients with ESRD using biopsy and necropsy procedures. The extent of renal fibrosis was histopathologically examined by Masson trichrome staining. T-cell and B-cell localization and macrophage lineages were determined by immunohistochemical staining. Additionally, interleukin-1 (IL-1), IL-2, and IL-6 levels in the canine ESRD kidney were immunohistochemically evaluated and compared with expression patterns in the normal kidney. Significant fibrosis and infiltrating immunocytes consistent with lymphocytes were observed. Although the B-cell count was increased in the end-stage kidney, immunostaining patterns disclosed a marked increase in the number of CD3(+) cells. Furthermore, the remarkable increase in IL-1 and IL-6 levels suggests that T cells in the kidneys of dogs with ESRD spontaneously express these cytokines. In this study, the correlation between the degree of renal fibrosis and cytokines in canine ESRD was examined. The present study shows that T lymphocytes and IL-6 play important roles in renal fibrosis.