Search

Abstract Machine learning (ML) models have gained significant attention in a variety of applications, from computer vision to natural language processing, and are almost always based on big data. There are a growing number of applications and products with built-in machine learning models, and this is the area where software engineering, artificial intelligence and data science meet. The requirement for a system to operate in a real-world environment poses many challenges, such as how to design for wrong predictions the model may make; How to assure safety and security despite possible mistakes; which qualities matter beyond a model’s prediction accuracy; How can we identify and measure important quality requirements, including learning and inference latency, scalability, explainability, fairness, privacy, robustness, and safety. It has become crucial to test thoroughly these models to assess their capabilities and potential errors. Existing software testing methods have been adapted and refined to discover faults in machine learning and deep learning models. This paper covers a taxonomy, a methodologically uniform presentation of all presented solutions to the aforementioned issues, as well as conclusions about possible future development trends. The main contributions of this paper are a classification that closely follows the structure of the ML-pipeline, a precisely defined role of each team member within that pipeline, an overview of trends and challenges in the combination of ML and big data analytics, with uses in the domains of industry and education.

Abstract This paper explores the potential of a multidisciplinary approach to testing and aligning artificial intelligence (AI), specifically focusing on large language models (LLMs). Due to the rapid development and wide application of LLMs, challenges such as ethical alignment, controllability, and predictability of these models emerged as global risks. This study investigates an innovative simulation-based multi-agent system within a virtual reality framework that replicates the real-world environment. The framework is populated by automated 'digital citizens,' simulating complex social structures and interactions to examine and optimize AI. Application of various theories from the fields of sociology, social psychology, computer science, physics, biology, and economics demonstrates the possibility of a more human-aligned and socially responsible AI. The purpose of such a digital environment is to provide a dynamic platform where advanced AI agents can interact and make independent decisions, thereby mimicking realistic scenarios. The actors in this digital city, operated by the LLMs, serve as the primary agents, exhibiting high degrees of autonomy. While this approach shows immense potential, there are notable challenges and limitations, most significantly the unpredictable nature of real-world social dynamics. This research endeavors to contribute to the development and refinement of AI, emphasizing the integration of social, ethical, and theoretical dimensions for future research.

Removal of Cu(II) ions from water solutions by sulphuric acid treated Lagenaria vulgaris shell (ccLVB) was studied. Batch experiments were done by shaking a fixed mass of biosorbent (1.0 g) with 250 cm3 of 50.0 mg dm–3 Cu(II) solutions, at pH ranged from 2 up to 6. Metal concentration in the filtrates as well as in the initial solution was determined by flame atomic absorption spectrometry. Results show that efficiency of Cu(II) ions uptake by sulphuric acid treated Lagenaria vulgaris shell is significantly greater than raw Lagenaria vulgaris biosorbent. In addition, there is no significant effect of initial pH of solution on Cu(II) ions uptake by ccLVB and obtained biosorbent can be applied in a wide range of pH.

The content and distribution of 21 metals in the central and peripheral parts of the foliose epiphytic lichen Flavoparmelia caperata (L.) Hale, collected in the area of Southeastern Serbia, were analysed in terms of biological monitoring. Inductively coupled plasma atomic emission spectrometry revealed higher concentrations of As, B, Ba, Cd, Ga, Pb, Se, Cr, Cu, Fe, In, Li and/or Ni in peripheral, younger, parts and Ba, K, Tl, Mg, Na and/or Zn in central, older parts of lichens. Principal component analysis and hierarchical cluster analysis were used to identify the relationship among metals in samples and their possible sources. Significant correlations were found among Ni-Cr, Cd-Ga-In-As-Se, Zn-Ba, Cu-Pb-B, suggesting a common source of pollution. Given the location of sampling, these findings probably reflect airborne metal pollution in relation to the main wind directions and vicinity of the roads and industrial complexes. The importance of this study is the evidence that the Special Nature Reserve Jelašnička Gorge is influenced by pollution sources in the area. Flavoparmelia caperata could be effective as an early indicator of environmental changes of the studied area.

As large language models (LLMs) continue to gain popularity due to their human-like traits and the intimacy they offer to users, their societal impact inevitably expands. This leads to the rising necessity for comprehensive studies to fully understand LLMs and reveal their potential opportunities, drawbacks and overall societal impact. With that in mind, this research conducted an extensive investigation into seven LLMs, aiming to assess the temporal stability and inter-rater agreement on their responses on personality instruments in two time points. In addition, LLMs’ personality profile was analysed and compared with human normative data. The findings revealed varying levels of inter-rater agreement in the LLMs’ responses over a short time, with some LLMs showing higher agreement (e.g. Llama3 and GPT-4o) compared with others (e.g. GPT-4 and Gemini). Furthermore, agreement depended on used instruments as well as on domain or trait. This implies the variable robustness in LLMs’ ability to reliably simulate stable personality characteristics. In the case of scales which showed at least fair agreement, LLMs displayed mostly a socially desirable profile in both agentic and communal domains, as well as a prosocial personality profile reflected in higher agreeableness and conscientiousness and lower Machiavellianism. Exhibiting temporal stability and coherent responses on personality traits is crucial for AI systems due to their societal impact and AI safety concerns.

Acute B. canis infection can lead to an acute phase reaction (APR) in dogs. The parasite invades red blood cells causing anemia through immune-mediated hemolysis and possible erythropoietic suppression. A regenerative response of the erythroid lineage during the babesiosis has not been described in extension. This research examines hematologic parameters focusing on the absolute reticulocyte count and apolipoprotein A I (ApoA I) level on the day of admission and 14 days after treatment with imidocarb-dipropionate in young (n=11) and adult (n=11) dogs naturally infected with B. canis. Metabolic and inflammatory processes were characterized by analyzing protein and lipid profiles, as well as ApoA I at specified time points. Automated analyzers were used to determine complete blood count and biochemical parameters, while ApoA I was assessed using radioimmunoassay. The reticulocyte count was determined using a manual method by means of supravital staining. Both young and adult dogs with acute B. canis infection showed non-regenerative anemia without difference. Fourteen days after successful treatment with imidocarb dipropionate, the anemia was corrected and a high reticulocyte count was observed (p

The 2022 release of ChatGPT sparked widespread discussions on Artificial General Intelligence (AGI). Through a detailed examination of an interview with Bard, a large language model, the study uncovers narratives of optimism and pessimism regarding AGI's future implications. It found a higher leaning towards optimism about AGI's potential effects, extending from education, arts, and relationships, to economy and space exploration. Conversely, pessimistic views pointed out potential downsides, such as unemployment, political instability, and media manipulation. The study also identified four primary AGI themes - the relationship between humans and AGI, AGI acquiring a physical form, AGI simulating a universe, and the responsible utilization of AGI. These insights aid in understanding the complex socio-technical imaginary surrounding AGI. The study has its limitations as it is based solely on the responses provided by Bard during its test phase. Additional research may reveal changes in AGI discourse representation as the model evolves.

Abstract Human umbilical artery (HUA) preparations are of particular importance for in vitro studies on isolated blood vessels because their sampling is not risky for the patient, and they can provide the closest possible impression of changes related to the uteroplacental circulation during pre‐eclampsia. Using organ bath techniques, useful experimental protocols are provided for measuring some pathophysiological phenomena in the vascular responses of HUAs. Several vasoconstrictors (serotonin, prostaglandin F and phenylephrine) and vasodilators (acetylcholine and minoxidil) were seleted for determination of their vasoactivity in HUAs. The role of L‐type voltage‐operated calcium channels and different types of potassium channels (KATP, BKCa and KV) were assessed, as was the impact of homocysteine. Serotonin was confirmed to be the most potent vasoconstrictor, while acetylcholine and phenylephrine caused variability in the relaxation and contraction response of HUA, respectively. The observed increase in serotonin‐induced contraction and a decrease in minoxidil‐induced relaxation in the presence of homocysteine suggested its procontractile effect on HUA preparations. Using selective blockers, it was determined that KATP and KV channels participate in the minoxidil‐induced relaxation, while L‐type voltage‐dependent Ca2+ channels play an important role in the serotonin‐induced contraction. The presented protocols reveal some of the methodological challenges related to HUA preparations and indicate potential outcomes in interpreting the vascular effects of the investigated substances, both in physiological conditions and in the homocysteine‐induced pre‐eclampsia model.

The InSAR technique is a powerful tool for monitoring land subsidence, which has the potential to provide millimeter-accurate displacement determinations. By using the free Sentinel-1 images, which have a very large spatial coverage and a high frequency of repeated passes over the same area, the InSAR technique can offer good spatial and temporal displacement analysis with high accuracy. Considering the existence of a significant number of free research software, and the need to engage a smaller number of professional staff, this technique is significantly more financially favorable than other methods.

Background and Objectives: Left ventricular aneurysm (LVA) is associated with a decline in cardiac function, evidenced by a lower ejection fraction (EF), due to the reduction in the proportion of functional myocardium. The left ventricular end-diastolic volume (LVEDV), the left ventricular aneurysm volume (LVAV), and the LVAV/LVEDV ratio show a strong correlation with the EF. The aim of this study was to determine LVA characteristics post-myocardial infarction (basal vs. apical) and to evaluate the impact of aneurysm volume in diastole (LVAVd), aneurysm area in diastole (LVAAd), and their respective ratios with LVEDV and area (LVEDA) on the EF, in order to identify the most critical predictive factors for assessing and managing the negative impact of aneurysms on cardiac function. Materials and Methods: This observational study included post-infarction LVA patients at the “Dedinje” Cardiovascular Institute in Belgrade, Serbia, undergoing routine transthoracic echocardiography. Echocardiography assessed volumes (LVEDV, LVESV, LVAVd, LVAVs) and areas (LVAAd, LVAAs, LVEDA, LVESA) using the area–length method. The ratios (LVAVd/LVEDV, LVAVs/LVESV, LVAAd/LVEDA, LVAAs/LVESA) were derived from these measures. The left ventricular EF was calculated using Simpson’s method. Results: Basal aneurysms showed a significantly smaller LVAVd (p = 0.016), LVAAd (p = 0.003), and LVAAs (p = 0.029) compared to apical aneurysms, indicating that basal aneurysms are smaller in size. However, there was no significant difference in the EF and overall LV volumes between the groups, although the basal aneurysm group had a slightly higher EF and end-diastolic volume, with a slightly lower end-systolic volume. Furthermore, when comparing the correlation between the EF and the LVAVd, the LVEDV, and the LVAVd/LVEDV ratio, the results indicate that the LVAVd had the greatest impact on the EF (−0.695), followed by the LVAVd/LVEDV ratio (−0.637), and the lowest correlation is between the EF and LVEDV. A similar relationship is observed when comparing the EF with the LVESV, the LVAVs, and the LVAVs/LVESV ratio. Conclusions: Basal aneurysms are significantly smaller than apical ones, yet EF and LV volumes remain similar between the groups, with the EF being slightly higher in the basal group. In cases of LVA, LVAVd shows the strongest negative correlation with the EF, indicating its significant impact on systolic function, followed by the LVAVd/LVEDV ratio, with the weakest correlation seen between the EF and LVEDV.

Patients suffering from cholelithiasis have an increased risk of developing cardiovascular complications, particularly ischemic myocardial disease. Ursodeoxycholic acid (UDCA), already used in clinical practice for the treatment of cholelithiasis and related conditions, has proven antioxidative, anti-inflammatory, and cytoprotective effects. Therefore, the aim of this study was to investigate the cardioprotective effect of UDCA pre-treatment on isoprenaline-induced myocardial injury in rats. Male Wistar albino rats were randomized into four groups. Animals were pre-treated for 10 days with propylene glycol + saline on days 9 and 10 (control), 10 days with propylene glycol + isoprenaline on days 9 and 10 (I group), 10 days with UDCA + saline on days 9 and 10 (UDCA group), and 10 days with UDCA + isoprenaline on days 9 and 10 (UDCA + I group). UDCA pre-treatment significantly reduced values of high-sensitivity troponin I (hsTnI) and aspartate aminotransferase (AST) cardiac markers (p < 0.001 and p < 0.01, respectively). The value of thiobarbituric acid reactive substances (TBARS) was also decreased in the UDCA + I group compared to the I group (p < 0.001). UDCA also significantly increased glutathione (GSH) levels, while showing a tendency to increase levels of superoxide dismutase (SOD) and catalase (CAT). The level of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) expression, a key regulatory gene of inflammation, was diminished when UDCA was administered. A reduction of cardiac damage was also observed in the UDCA pre-treated group. In conclusion, UDCA pre-treatment showed a cardioprotective effect on isoprenaline-induced myocardial injury in rats, primarily by reducing oxidative stress and inflammation.

Galectins are a class of lectins that are extensively expressed in all organisms. Galectins are involved in a range of functions, including early development, tissue regeneration, cancer and inflammation. It has been shown that galectin-8 is expressed in the villous and extravillous trophoblast (EVT) cells of the human placenta; however, its physiological role in pregnancy establishment has not been elucidated. Taking these factors into account, we investigated the functional role of galectin-8 in HTR-8/SVneo cells—a human EVT cell line—and human primary cytotrophoblast cells isolated from a first-trimester placenta. We analyzed the effects of recombinant human galectin-8 (rh galectin-8) on the adhesion, migration and invasion of HTR-8/SVneo cells. We used qPCR, cell-based ELISA (cELISA) and gelatin zymography to study the effects of galectin-8 on mediators of these processes, such as integrin subunits alpha-1 and beta-1 and matrix metalloproteinases (MMPs)-2 and -9, on the mRNA and protein levels. Further, we studied the effects of galectin-8 on primary cytotrophoblast cells’ invasion. Galectin-8 stimulated the adhesion, migration and invasion of HTR-8/SVneo cells, as well as the invasion of primary cytotrophoblasts. In addition, the MMP-2 and -9 levels were increased, while the expression of integrins alpha-1 and beta-1 was not affected. Galectin-8 has the ability to positively affect EVTs’ invasion, so it can be considered a significant factor in the trophoblast cell invasion process.

Background and Objectives: The saphenous vein graft (SVG) remains the most frequently used conduit worldwide, despite its common disadvantage of early graft failure. To solve the problem and reduce the SVG damage, Souza implemented a new technique where a vein is harvested with surrounding fascia and fat tissue (the so-called no-touch technique). Materials and Methods. A prospective study conducted from February 2019 to June 2024 included 23 patients who underwent myocardial revascularization using a no-touch vein, with follow-up control examinations using computed tomographic angiography to detect graft stenosis or occlusion. Results. Of the entire patient group, 17 (73.9%) were male, with a mean age of 67.39 ± 7.71 years. The mean follow-up period was 25 months. There were no major adverse cardiovascular or cerebrovascular events (MACCEs) during hospitalization, although one patient died in the hospital. Another patient died due to malignancy, but no MACCEs occurred during the follow-up period. According to multi-slice CT coronary angiography, the results were impeccable, with an astonishing 100% patency observed in all 20 IMA grafts and 58 no-touch SVGs examined. Conclusions. The excellent patency rate during the early follow-up period confirmed that the no-touch technique is a good option for surgical revascularization.

Background and Objectives. Distinct pressure curve differences exist between akinetic (A-LVA) and dyskinetic (D-LVA) aneurysms. In D-LVA, left ventricular (LV) ejection pressure decreases relative to the aneurysm size, whereas A-LVA does not impact pressure curves, indicating that the decrease in stroke volume (SV) and cardiac output is proportional to the size of dyskinesia. This study aimed to assess the frequency of A-LVA and D-LVA, determine aneurysm size parameters (volume and surface area), and evaluate predictive parameters using echocardiography in A-LVA and D-LVA. Furthermore, it aimed to compare individual echocardiographic parameters, according to ejection fraction (EF) and SV, with hemodynamic events shown in experimental models of A-LVA and D-LVA and their significance in everyday clinical practice. Materials and Methods. This clinical study included patients with post-infarction left ventricular aneurysm (LVA) admitted to the cardiovascular institute ‘’Dedinje”, Serbia. Echocardiographic volume and surface area of LV and LVA were determined (by the area–length method) along with EF (by Simpson’s method). Results. A-LVA was present in 62.9% of patients, while D-LVA was present in 37.1%. Patients with D-LVA had significantly higher systolic aneurysm volume (LVAVs) (94.07 ± 74.66 vs. 51.54 ± 53.09, p = 0.009), systolic aneurysm surface area (LVAAs) (23.22 ± 11.73 vs. 16.41 ± 8.58, p = 0.018), and end-systolic left ventricular surface areas (LVESA) (50.79 ± 13.33 vs. 42.76 ± 14.11, p = 0.045) compared to patients with A-LVA. The ratio of LVA volume to LV volume was higher in the D-LVA in systole (LVAVs/LVESV). The end-diastolic volume of LV (LVEDV) and end-systolic volume of LV (LVESV) did not significantly differ between D-LVA and A-LVA. EF (21.25 ± 11.92 vs. 28.18 ± 11.91, p = 0.044) was significantly lower among patients with D-LVA. Conclusions. Differentiating between A-LVA and D-LVA using echocardiography is crucial since D-LVA causes greater hemodynamic disturbances in LV function, and thus surgical resection of the aneurysm or LV reconstruction must have a positive effect regardless of myocardial revascularization surgery.

Considering the escalating global prevalence and the huge therapeutic demand for the treatment of hypertension, there is a persistent need to identify novel target sites for vasodilator action. This study aimed to investigate the role of TRPA1 channels in carvacrol-induced vasodilation and to design novel compounds based on carvacrol structure with improved activities. In an isolated tissue bath experiment, it was shown that 1 µM of the selective TRPA1 antagonist A967079 significantly (p < 0.001) reduced vasodilation induced by 3 mM of carvacrol. A reliable 3D-QSAR model with good statistical parameters was created (R2 = 0.83; Q2 = 0.59 and Rpred2 = 0.84) using 29 TRPA1 agonists. Obtained results from this model were used for the design of novel TRPA1 activators, and to predict their activity against TRPA1. Predicted pEC50 activities of these molecules range between 4.996 to 5.235 compared to experimental pEC50 of 4.77 for carvacrol. Molecular docking studies showed that designed molecules interact with similar amino acid residues of the TRPA1 channel as carvacrol, with eight compounds showing lower binding energies. In conclusion, carvacrol-induced vasodilation is partly mediated by the activation of TRPA1 channels. Combining different in silico approaches pointed out that the molecule D27 (2-[2-(hydroxymethyl)-4-methylphenyl]acetamide) is the best candidate for further synthesis and experimental evaluation in in vitro conditions.