Traffic accidents are very serious problems for human life and the environment. In road safety, it is crucial to identify the high risk locations to apply proper counter measures. This paper aims at introducing outcomes of a pilot project whose main goal is to develop a GIS based crash analysis system integrated with the quantitative methods for identification of high risk locations on road networks in Turkey. In this concept, traditional hotspot detection methods used in Turkey(crash frequency, rate, and severity) are compared with the spatial statistical methods including Moran's I, GetisOrd G and planar and network kernel density estimation in terms of their sensitivity to spatial characteristics of crash clusters. Many countries use traditional hotspot detection approaches such as crash frequency, crash rate, and crash severity as well as Turkey. In this project, we aimed at obtaining a model including different hotspot identification methods for the safety program of Turkey. In order to obtain the model, many hotspot detection methods will be used and compare stage by stage. In the first stage, the seven methods mentioned above are used and examined. Although some of these methods are compared in couple, there is no study using all these methods together extensively in the literature. Methods validated with a different spatial vantage points. Repetitiveness of hotspots in a seven years period are used to compare the methods. Meanwhile advantages and disadvantages of the methods according to location of hotspots are examined additionally. Results show that using planar KDE with Gi in the junction locations and using planar KDE with Moran's I in the straight road locations could improve the model while determining hotspots.

Lung cancer is the leading cancer type that causes mortality in both men and women. Computer-aided detection (CAD) and diagnosis systems can play a very important role for helping physicians with cancer treatments. This study proposes a hierarchical deep-fusion learning scheme in a CAD framework for the detection of nodules from computed tomography (CT) scans. In the proposed hierarchical approach, a decision is made at each level individually employing the decisions from the previous level. Further, individual decisions are computed for several perspectives of a volume of interest. This study explores three different approaches to obtain decisions in a hierarchical fashion. The first model utilizes raw images. The second model uses a single type of feature image having salient content. The last model employs multi-type feature images. All models learn the parameters by means of supervised learning. The proposed CAD frameworks are tested using lung CT scans from the LIDC/IDRI database. The experimental results showed that the proposed multi-perspective hierarchical fusion approach significantly improves the performance of the classification. The proposed hierarchical deep-fusion learning model achieved a sensitivity of 95% with only 0.4 fp/scan.