Automatic topic Detection seems unavoidable in social media analysis due to big text data which their users generate. Clustering-based methods are one of the most important and up-to-date categories in topic Detection. The goal of this research is to have a wide study on this category. Therefore, this paper aims to study the main components of clustering-based-topic-Detection, which are embedding methods, distance metrics, and clustering algorithms. Transfer learning and consequently pretrained language models and word embeddings have been considered in recent years. Regarding the importance of embedding methods, the efficiency of five new embedding methods, from earlier to recent ones, are compared in this paper. To conduct our study, two commonly used distance metrics, in addition to five important clustering algorithms in the field of topic Detection, are implemented by the authors. As COVID-19 has turned into a hot trending topic on social networks in recent years, a dataset including one-month tweets collected with COVID-19-related hashtags is used for this study. More than 7500 experiments are performed to determine tunable parameters. Then all combinations of embedding methods, distance metrics and clustering algorithms (50 combinations) are evaluated using Silhouette metric. Results show that T5 strongly outperforms other embedding methods, cosine distance is weakly better than other distance metrics, and DBSCAN is superior to other clustering algorithms.

Introduction. Ultrasonography is the preferable imaging technique for monitoring and assessing complications in kidney allograft transplants. Computer-Aided diagnostic system based on texture analysis in ultrasonographic imaging is recommended to identify changes in kidney function after allograft transplantation. Materials and Methods. A total of 61 biopsy-proven kidney allograft recipients (11 rejected and 50 unrejected) were assessed by a Computer-Aided diagnostic system. Up to 270 statistical texture features were extracted as descriptors for each region of interest in each recipient. Correlations of texture features with serum creatinine level and differences between rejected and unrejected allografts were analyzed. An area under the receiver operating characteristic curve was calculated for each significant texture feature. Linear discriminant analysis was employed to analyze significant features and increase discriminative power. Recipients were classified by the first nearest neighbor classifier. Results. Fourteen texture features had a significant correlation with serum creatinine level and 16 were significantly different between the rejected and unrejected allografts, for which an area under the curve values were in the range of 0. 575 for difference entropy S(4, 0) to 0. 676 for kurtosis. Using all 16 features, linear discriminant analysis indicated higher performance for classification of the two groups with an area under the curve of 0. 975, which corresponded to a sensitivity of 90. 9%, a specificity of 100%, a positive predictive value of 100%, and a negative predictive value of 98. 0%. Conclusions. Texture analysis was a reliable method, with the potential for characterization, and can help physicians to diagnose kidney failure after transplantation on ultrasonographic imaging.

INTRODUCTIONExtraction and processing of various features with the help of aerial imagery reduces the time and financial costs associated with the use of ground mapping and the resulting human error. Advances in the field of aerial sensors in terms of spatial and spectral resolution with precise place and performance picking up altitude from the ground have led to the use of each part of information about terrestrial phenomena such as spectral and spatial characteristics Brought. Today, complementary data used to detect complications are Lidar data, the sensor of which is sent and received, and the electromagnetic spectrum in the near-infrared spectrum (in its aerial form) and joined the spectrum. Pays close infrared and green band (in space type). DATA AND METHODSLidar data and spectral images were analyzed using different types of algorithms effective in landfill extraction to assess density. New layers of images were obtained in the form of raster from the study area, which was analyzed by performing slope extraction steps on flat and sloping surfaces. Buildings that were definitely not buildings were removed. The size and spectral characteristics of the missing structures were identified and the parcels were redistributed to extract the impermeable surfaces. Which led to the achievement of two levels of parcels and impenetrable points. The data set is related to the northern part of Bandar Anzali, which includes a vertical aerial photograph, irregular cloud points of the region with dense one to two points per square meter with an average point space of 0.69 square meters, and vertical aerial photograph with spatial resolution. It is 8 cm square. RESULTS AND DISCUSSIONIn this study, a different method for extracting buildings using airborne Lidar data and ultracam images was presented. The proposed system used geometric and spatial information of Lidar data and ultracam images, which included three general steps, in the first step; Lidar data were filtered and extracted using spectral clustering of buildings. In the second step; The obtained model was compared with the two-dimensional boundaries of buildings by the height threshold method. In the third step; After extraction, the first building boundaries were merged with the structures extracted by the checker algorithm. In the stage of separating terrestrial from non-terrestrial points, all points related to land were classified and extracted. The remaining points were classified as roof points, which were dealt with in the fault section of the buildings. All the functions used enabled the system to successfully extract the structures from the Lidar data. CONCLUSIONThe data for the first return points were subtracted from the data for the last return points and a fixed value was obtained which depended on the altitude accuracy of the difference between the two returns. In addition to the mentioned method, the clustering method was used during the research that each cluster belonged to a roof section so that the characteristics of each surface model could be easily determined.Then, to complete the shape of the roof, the footprint of the building that was extracted was used. In fact, the borderlines and inner vertices extracted only part of the shape of the border. Other sections, such as vertical edges, were not detected due to intersection. This is due to the lack of front sampling. Finally, the items extracted through spectral clustering in eCoginition software and two-dimensional boundaries extracted from ENVI Lidar software, to increase the accuracy of land surface extraction (buildings) and make the area of ​​buildings studied in this data Were merged. As mentioned; After extraction, the primary building boundaries were merged with the structures extracted by the checker algorithm. In the section of buildings diagnostics, buildings with errors were discussed and the evaluation of the results showed that the system used has relatively reached the set goals and the methods used include the threshold method. Elevation, clustering, spectral method, and integration method were evaluated for each of the four blocks with error rates of 28%, 15%, and 0%, respectively, based on the area of ​​extracted tolls to the study area. The error of each building was first examined in general and then in detail, and it was found that aerial Lidar technology has an extraordinary ability to collect very right and dense samples of altitude measurements of cities and a new level of detail information can be Accurately extracted building density automatically and efficiently from aerial Lidar data. In 417 buildings that were surveyed and analyzed, the height of the buildings was achieved with high accuracy and all the buildings in the study area were extracted with a compact and organic density as well as scattered and planned.