Search

The paper presents the development of a method of computer-aided diagnosis of sarcoidosis based on chest X-ray images, for particular stages of the disease. For this purpose, the research material, which consisted of 98 chest X-rays, was analyzed. The datasets included images for healthy cases and for first, second and third degree sarcoidosis. The research material was pre-processed, after which, on the basis of framing, the regions of interest (ROIs) were extracted from the images for individual cases. Next, the analysis of the selected ROIs was carried out, resulting in discriminatory characteristics describing the properties of the images. For the obtained sets, due to their multidimensionality, extraction and selection of features were carried out. Based on the analysis of the obtained results, a selection of features was selected to reduce the data dimension. Three methods were used to carry it out. In the case of heuristic identification of variables, datasets counting respectively for set X-ray2: 34, X-ray3: 47 textural features were obtained. On the basis of the obtained sets, classifiers were built using the supervised learning method. As a result, one model was obtained, based on a single classifier, for the X-ray2 dataset, with a classification error equal to zero. For the X-ray3 dataset, one model was obtained, which was based on an aggregated classifier consisting of two component classifiers and for which the classification error was also equal to zero. The resulting models were proposed as a final solution. The resulting feature vectors and models obtained during the research can be used to build a computer system that will carry out the diagnostic process automatically. The developed solution allows us to classify images for X-ray imaging, depending on the degree of sarcoidosis, into two categories: healthy or sick. This makes it possible to build a system that improves the work of the diagnostician in the process of diagnosing the disease, by reducing the time and cost of performing image analysis, as well as for the patient's condition, thanks to faster referral to advanced clinical trials. [ABSTRACT FROM AUTHOR]

Facial recognition systems are widely used in various daily applications due to their high accuracy rate and ease of use. Unfortunately, their performance drops drastically when encountering spoofing attacks. In this paper, a novel pipeline for face spoofing detection based on the extraction of Heterogeneous Auto-Similarities of Characteristics (HASC) descriptor from the HSV and YCbCr color spaces is proposed. To detect print and replay spoofing attacks, the HASC descriptor encodes linear and non-linear dependencies between low-level dense features of the color face image using covariance and information-theoretic measures. The combination of covariance and information-theoretic measures makes HASC an efficient descriptor to explore how the features are interrelated in real and fake faces. The HASC texture descriptions are extracted separately from each color channel of the HSV and YCbCr color spaces, and the resulting feature vectors are concatenated into an extended feature vector to better discriminate between real and fake faces. Extensive experiments are conducted on publicly available databases namely, Replay-Attack, CASIA FASD, and MSU MFSD. The obtained results demonstrate that while being of low dimension, the HASC descriptor when fed into a binary Support Vector Machine (SVM) classifier, achieves competitive results compared to the state-of-the-art. [ABSTRACT FROM AUTHOR]

• Surface roughness is a complex multiscale characteristic and multiple roughness indexes can be defined. • The simplified geostatistical-based algorithm presented describes key aspects of short-range roughness. • Roughness anisotropy is a key morphological characteristic. Surface roughness, i.e. local variations in elevation which indicate less and less local smoothness, is an important concept for the analysis of solid earth/planetary surfaces and of the related surface processes. Surface roughness is a general concept covering multiple aspects of the surface spatial variability structure, that can be characterized at different scales and with different metrics. Accordingly, it is not surprising the wide range of approaches and algorithms adopted for the analysis of surface roughness and the lack of an accepted standard for its evaluation. In this context, geostatistical-based roughness indexes are a valid solution, providing a good balance between flexibility of algorithms and interpretability of the results. However, despite the long record of applications and the well-known and robust theoretical framework, geostatistical-based surface roughness tools have still not gained momentum in the context of geomorphometric analysis. Many geomorphometric studies concerning with the evaluation of roughness indexes are based on popular approaches, such as the ones based on measures of dispersion of normal vectors to surface. Unfortunately, these indexes present some drawbacks in the capability to represent specific aspects of surface roughness and from the perspective of their interpretability. Accordingly, in this paper some of the key aspects and advantages of geostatistical-based approaches for roughness analysis are highlighted, including the relevance of roughness anisotropy. Then, a new approach tailored to the analysis of short-range roughness, requiring a minimum intervention by the user, is introduced discussing the advantages compared to the vector dispersion-based roughness. The new approach is implemented as open-source code both in R language using the functions of the "Terra" package as well as in Python for Esri ArcMap GIS. The methodological considerations and algorithms presented can be applied also in the broader context of image analysis. [ABSTRACT FROM AUTHOR]