An Explainable Approach Based on Emotion and Sentiment Features for Detecting People with Mental Disorders on Social Networks

Mental disorders are a global problem that widely affects different segments of the population. Diagnosis and treatment are difficult to obtain, as there are not enough specialists on the matter, and mental health is not yet a common topic among the population. The computer science field has proposed some solutions to detect the risk of depression, based on language use and data obtained through social media. These solutions are mainly focused on objective features, such as n-grams and lexicons, which are complicated to be understood by experts in the application area. Hence, in this paper, we propose a contrast pattern-based classifier to detect depression by using a new data representation based only on emotion and sentiment analysis extracted from posts on social media. Our proposed feature representation contains 28 different features, which are more understandable by specialists than other proposed representations. Our feature representation jointly with a contrast pattern-based classifier has obtained better classification results than five other combinations of features and classifiers reported in the literature. Our proposal statistically outperformed the Random Forest, Naive Bayes, and AdaBoost classifiers using the parser-tree, VAD (Valence, Arousal, and Dominance) and Topics, and Bag of Words (BOW) representations. It obtained similar statistical results to the logistic regression models using the Ensemble of BOWs and Handcrafted features representations. In all cases, our proposal was able to provide an explanation close to the language of experts, due to the mined contrast patterns.

An Explainable Artificial Intelligence Model for Detecting Xenophobic Tweets

Xenophobia is a social and political behavior that has been present in our societies since the beginning of humanity. The feeling of hatred, fear, or resentment is present before people from different communities from ours. With the rise of social networks like Twitter, hate speeches were swift because of the pseudo feeling of anonymity that these platforms provide. Sometimes this violent behavior on social networks that begins as threats or insults to third parties breaks the Internet barriers to become an act of real physical violence. Hence, this proposal aims to correctly classify xenophobic posts on social networks, specifically on Twitter. In addition, we collected a xenophobic tweets database from which we also extracted new features by using a Natural Language Processing (NLP) approach. Then, we provide an Explainable Artificial Intelligence (XAI) model, allowing us to understand better why a post is considered xenophobic. Consequently, we provide a set of contrast patterns describing xenophobic tweets, which could help decision-makers prevent acts of violence caused by xenophobic posts on Twitter. Finally, our interpretable results based on our new feature representation approach jointly with a contrast pattern-based classifier obtain similar classification results than other feature representations jointly with prominent machine learning classifiers, which are not easy to understand by an expert in the application area.

Computed tomographic pneumogastrography in determining the types of gastric cancer according to the Lauren classification

Objective. To assess the capabilities of computed tomographic pneumogastrography in determining the types of gastric cancer according to the Lauren classification at the stage of clinical staging.Materials and methods. This study is a single-center retrospective study with 202 patients with gastric cancer included who was treated at the National Medical Research Center of Oncology named after N. N. Petrov from 2015 to 2018. All patients underwent subtotal gastric resection or gastrectomy and computed tomographic pneumogastrography at the stage of clinical staging. For patients undergoing neoadjuvant chemotherapy, CT was performed twice: before chemotherapy and after, immediately before surgery. We studied quantitative and qualitative imaging biomarkers, measured densitometric indices of stomach tumor density in the arterial, portal and delayed phases of scanning at five different points. For patients receiving NACT, all density indices were recorded twice — both before the start of therapeutic treatment, and after, immediately before the surgery.Results. The distribution of gastric cancer types according to Lauren»s classification was as follows: in 59 (29,2 %) intestinal type, 69 (34,2 %) — diffuse, 16 (7,9 %) — mixed, 58 (28,7 %) — indeterminate type. Based on visual characteristics, taking into account the characteristics of tumor growth, 3 main CT-PGG of the gastric cancer type were identified: 1 — tuberous (n = 68, 34,0 %), 2 — intramural (n = 57,3 %) and 3 — mixed (n = 77,4 %). CT-PGG tumor type is associated with Lauren type (χ2 = 185,19, p <0,001). With a tuberous CT-PGG type, it is possible to assume that the tumor is of an intestinal or indeterminate Lauren type; sensitivity 0,58 (95% CI: 0,49-0,67), specificity 0,1 (95% CI: 0,96-0,1). With an intramural CT-PGG type, the diffuse type of tumor according to Lauren is most likely; sensitivity 0,75 (95% CI: 0,64-0,85), specificity 0,96 (95% CI: 0,91-0,99). With a mixed CT-PGG type, the definition of the type according to Lauren is difficult. In the definition of mixed tumor type according to Lauren, the sensitivity and specificity of mixed CT-PGG tumor type are 0,94 (95% CI: 0,70% -0,1) and 0,67 (95% CI: 0,59-0,73) respectively.Conclusion. The shape of the stomach tumor, determined by CT-PGG, has a high diagnostic efficiency in determining the types of gastric cancer according to Lauren. The tuberous CT-PGG type is typical for tumors of the intestinal type according to Lauren, and the intramural CT-PGG type is typical for tumors of the diffuse type according to Lauren. Tumors of indeterminate Lauren type have any CT-PGG type and contrast pattern. For tumors of a mixed type according to Lauren, a mixed type according to CT-PGG is characteristic, but differential diagnosis with tumors of a tuberous and diffuse type according to Lauren of an atypical form for them is impossible. Tumors of the intestinal and diffuse Lauren type of the CT-PGG type, which is not typical for them, have an atypical contrast pattern.

Diagnosis of epithelioid hemangioendothelioma of the liver. Lecture and analysis of own cases

Epithelioid hemangioendothelioma of the liver (EHEL) is a rare (1 case per 1 000,000 people) primary malignant neoplasm from the group of mesenchymal tumors. EHEL is the most aggressive representative of the hemangioendothelioma range with metastatic response and relapse rate, and therein is close to angiosarcoma.The purpose of this publication is to summarize our own experience in comparison with published data and to describe a disease symptom that was not investigated before.Materials and methods. We studied five cases of epithelioid hemangioendothelioma of the liver, confirmed by histological examination. All patients underwent standard computer tomography with bolus contrast agent.Results. The analysis of the images revealed typical for this pathology symptoms such as local retraction of the liver contour, a specific contrast pattern – “the lollipop sign”. We have revealed a new symptom: “the beading sign”, which was detected in four of five cases. In three of five cases, this symptom allowed us to determine the nature of the disease before histological examination.Conclusion. EHEL is a difficult disease to diagnose due to its rare frequency and similarity to other focal liver lesions. In aspect of radiology, it is advisable to focus on specific symptoms of this pathology: retraction of the contour, “lollipop-sign” and “beading-sign”.

Modelling and modification of simultaneous lightness contrast effect using the Albers' pattern

Experiments were carried out to investigate the simultaneous lightness contrast effect on a self-luminous display using simultaneous colour matching method. The Albers ' contrast pattern named ' double-crosses ' was used. The goals of this study were to model lightness contrast effect and modify it in the CAM16 colour appearance model. Five coloured targets were studied, and 41 test/background combinations were displayed on a calibrated display. Twenty normal colour vision observers performed colour matching in the experiment. In total, 820 matches were accumulated. The result shows present CAM16 has an unsatisfactory prediction for the effect, especially in the positive region which means the background is brighter than the target. Two models were established based on the visual data, i. e., with and without modification to the lightness difference in CAM16 space. Both of the models predict the effect with high accuracy and reliability.