The de.NBI cloud federation in 2025

We aimed to develop and evaluate Explainable Artificial Intelligence (XAI) for fetal ultrasound using actionable concepts as feedback to end-users, using a prospective cross-center, multi-level approach. We developed, implemented, and tested a deep-learning model for fetal growth scans using both retrospective and prospective data. We used a modified Progressive Concept Bottleneck Model with pre-established clinical concepts as explanations (feedback on image optimization and presence of anatomical landmarks) as well as segmentations (outlining anatomical landmarks). The model was evaluated prospectively by assessing the following: the model’s ability to assess standard plane quality, the correctness of explanations, the clinical usefulness of explanations, and the model’s ability to discriminate between different levels of expertise among clinicians. We used 9352 annotated images for model development and 100 videos for prospective evaluation. Overall classification accuracy was 96.3%. The model’s performance in assessing standard plane quality was on par with that of clinicians. Agreement between model segmentations and explanations provided by expert clinicians was found in 83.3% and 74.2% of cases, respectively. A panel of clinicians evaluated segmentations as useful in 72.4% of cases and explanations as useful in 75.0% of cases. Finally, the model reliably discriminated between the performances of clinicians with different levels of experience (p- values < 0.01 for all measures) Our study has successfully developed an Explainable AI model for real-time feedback to clinicians performing fetal growth scans. This work contributes to the existing literature by addressing the gap in the clinical validation of Explainable AI models within fetal medicine, emphasizing the importance of multi-level, cross-institutional, and prospective evaluation with clinician end-users. The prospective clinical validation uncovered challenges and opportunities that could not have been anticipated if we had only focused on retrospective development and validation, such as leveraging AI to gauge operator competence in fetal ultrasound.

Categorization of Emergent Processes by Students at Different Levels of Expertise Randi A. Engle (RAEngle@pitt.edu) Michelene T. H. Chi (chi@pitt.edu) Learning Research and Development Center, University of Pittsburgh 3939 O’Hara St., Pittsburgh, PA 15260 USA Introduction Chi and Roscoe (2002) proposed that one reason for the persistence of scientific misconceptions is that students classify scientific phenomena into incorrect ontological categories. In particular, Chi (submitted) has hypothesized that many commonly misunderstood science concepts (like evolution and electric current) are emergent processes in which a macro-level phenomena emerges from complex interactions between entities at a micro-level. Rather than correctly categorizing them this way, students are thought to miscategorize them as non-emergent processes in which there is a more direct relationship between the two levels. To test this hypothesis, we are conducting a study comparing how participants at different levels of expertise categorize science problems across domains, using a card sorting task modeled on Chi, Feltovich & Glaser (1981). The prediction of Chi’s theory is that participants with more expertise will be more likely to distinguish emergent from non-emergent problems while those with less expertise will often conflate them. Experts will also be more likely to refer to the 11 features of the emergent schema (e.g., disjointness, parallelism; see Chi, submitted) in defining their categories. Method Participants Participants consisted of 10 undergraduate and 9 doctoral students from the biology, chemistry, or physics departments at a local university. Undergraduates had all completed 1 st year courses in biology, chemistry, and physics. Eight were single majors in one of these disciplines and two were double majors in biology and chemistry. Doctoral students were in their third year or above. Materials Participants sorted 24 science problems, 8 drawn from each discipline’s 1 st year course. Within each discipline, half were emergent and half were non-emergent processes. Procedure Participants were asked to sort the problems into piles with similar mechanisms for linking the macro and micro levels. In the 1 st sort, they were allowed to make as many piles as they wished. In the 2 nd sort, they were asked to divide the cards into just two piles. In the 3 rd sort, the experimenter sorted the cards into emergent vs. non-emergent processes, and participants were asked to infer the distinction. In all cases, participants were asked to explain the explanatory mechanism each pile represented and why each problem fit. Results As a preliminary measure of the degree to which participants distinguished emergent from non-emergent processes, we calculated a weighted average (by pile size) of the absolute difference between the number of emergent versus non-emergent processes in each pile by the number of cards in the pile. For example, 3 piles—one with 5 emergent & 5 non-emergent problems, a 2 nd with 1 emergent & 7 non-emergent problems, and a 3 rd with 6 emergent & 0 non-emergent problems—would get a score of [10(0/10) + 8(6/8) + 6(6/6)] / 24 = 12/24 = .50. A value of 1 on this score represents perfect separation of emergent versus non-emergent processes while 0 represents perfect 50/50 mixing. Mean separation scores were .58 for doctoral students, .53 for single majors, and .73 for double majors, although none of these differences are statistically reliable. A complicating factor is that participants sometimes put, for example, only emergent problems in a given pile for reasons unrelated to their emergence. Thus, we analyzed participants’ definitions of their categories to determine whether they referred to any of the 11 features of the emergent schema. Doctoral students’ categories referred to more emergent features (1.78) than single majors (0.13; t(9) = 3.08, p < .05). Double majors referred to even more of them (3.50), but more data is needed to see if this is reliable. Discussion Doctoral students and double majors used more features of the emergent schema in sorting science problems than single majors, although they were no more likely than single majors to create piles that distinguished emergent and non- emergent processes. In future work, we will include in our sample more double majors as well as professors to further investigate the nature of expertise in emergent processes. References Chi, M. T. H. (submitted). “Causal” and “Emergent” Explanatory Mechanisms: Potential Schemas for Overcoming Misunderstandings in Science. Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science, 5, 121-152. Chi, M. T. H., & Roscoe, R. (2002). The processes and challenges of conceptual change. In M. Limon & L. Mason (Eds.), Reconsidering conceptual change: Issues in theory and practice. Dordrecht, NL: Kluwer.

Diabetes mellitus is not only a medical, but also a public issue. The number of people with diabetes is increasing every year. Diabetes causes complications of diseases. In addition to infertility issues and foot injuries that worsen the health condition, patients develop diabetic retinopathy, diabetic neuropathy, chronic kidney disease, cardiovascular diseases (strokes, ischemic heart diseases, peripheral vascular diseases). Accordingly, training courses for patients and focus on the health of the patients themselves are very important. The object of the research is training of persons with diabetes mellitus and skills favourable to health that help to prevent complications. The aim of the article is to reveal the process of learning of patients with diabetes and the application of skills favourable to health, which help to prevent complications in life. Tasks: 1) determine what kind of training courses are organized in Lithuania for persons with diabetes mellitus; 2) determine what kind of knowledge about health and disease control is provided by medical personnel; 3) reveal what kind of knowledge about health and what skills patients adapt in everyday life. Methods of research: theoretical – analysis of scientific literature, generalization; empirical – questionnaire survey, data analysis was performed by using the SPSS 23.0 program. The t-test was used to compare averages between groups, significant difference was considered to be p < 0,05. Data analysis was performed by applying calculations of non-parametric and parametric criteria. There were 61 patients with diabetes mellitus, aged from 13 to 78 years, who participated in the research. The research sought to reveal what training courses are organized for persons with diabetes mellitus and how patients themselves preserve their health. During the research, participants were questioned about the training courses they have attended; how patients with diabetes mellitus control their disease; and whether they follow the principles of a healthy lifestyle. Empirical research revealed that there are short-term training courses for patients with diabetes mellitus organized in Lithuania. The majority of participants noted that trainings are carried out for several hours or 7 days. Training courses that are carried out for patients with diabetes cover topics of a healthy lifestyles: balanced diet, physical activity, personal hygiene, blood sugar monitoring and insulin use; however, participants emphasize that such trainings should include more health aspects and that they prefer to have regular training courses. The research revealed that most training courses for patients with diabetes mellitus are carried out by a nurse, an endocrinologist and a dietician. Medical staff provides patients with diabetes with basic knowledge about health that includes a healthy lifestyle, foot care and recommendations for treatment (monitoring of blood suga,insulin use). It has been determined that patients with diabetes mellitus partially adapt the received knowledge. The research revealed that only 23.8 percent of participants with diabetes keep to a healthy diet plan, only 14.3 percent are physically active, and 39.7 percent care for their feet. Participants of the research lack knowledge about a healthy lifestyle and skills favourable to health that help to prevent complications. Only a small number of participants of the research with diabetes mellitus behave consciously and are prone to preserve their health by following recommendations of doctors. Therefore, patients with diabetes in Lithuania need regular and continuous training programs that would cover aspects of a healthy lifestyle (healthy and balanced diet, promotion of physical activity, personal hygiene). It is recommended to develop training programs that would include active training methods and videos, which would help patients with diabetes mellitus to more easily and independently acquire necessary health skills.

Assessing the reliability, validity and acceptance of a classification scheme of usability problems (CUP)

Triage is a complex process in ED that necessitates accurate, consistent and timely decision for assessment and management of incoming patients. With triage officers having different levels of expertise and experience, the triage scales are applied inconsistently, as evident from the lack of uniformity in the existing triaging system in Malaysian hospitals. In many hospitals, the triaging largely occurs in an ad hoc basis, placing patients at risk and contributing to inefficient resource utilisation. A computer system can be used to overcome these problems. Consequently, the paper proposes the development of a knowledge-based, agent-oriented decision support system for triage assessment and management. The ability to implement goal-directed behaviour, communicative processes and schedule-oriented, but concurrently executed tasks among others, make the agent technology suitable to be applied. We intend to adopt a modelling view to knowledge engineering, and plan to use the CommonKADS methodology as a basis for knowledge engineering, modelling and knowledge base development.

Interacting to reach a shared decision is an omnipresent component of human collaboration. We explored the interaction between dyads of individuals with different levels of expertise. The members of the dyads completed a number line task privately, jointly and privately again. In the joint condition, dyad members shared their private estimates and then negotiated a joint estimate. Both dyad members averaged their private individual estimates to determine joint estimates, thereby showing a strong equality bias. Their performance in the joint condition exceeded the performance of the dyad's best estimator, demonstrating interaction benefit, only when the dyad members had similar levels of expertise and when the averaged dyad performance was sufficiently accurate. At the end of the task, participants rated their level and their partner's level of competence. Participants were accurate in classifying themselves as the expert or the novice within the dyad. Nevertheless, novices tended to overestimate their ability as they admitted to being less competent but only slightly worse than their expert partner. Experts, instead, believed themselves to be more competent but were humble and considered their performance only marginally better than their partner's. Overall, these results have important implications for settings in which people with different levels of expertise interact. (PsycINFO Database Record

Training models in laparoscopic surgery allow the surgical team to practice procedures in a safe environment. We have proposed the use of a 4-task, low-cost inert model to practice critical steps of laparoscopic common bile duct exploration. The performance of 3 groups with different levels of expertise in laparoscopic surgery, novices (A), intermediates (B), and experts (C), was evaluated using a low-cost inert model in the following tasks: (1) intraoperative cholangiography catheter insertion, (2) transcystic exploration, (3) T-tube placement, and (4) choledochoscope management. Kruskal-Wallis and Mann-Whitney tests were used to identify differences among the groups. A total of 14 individuals were evaluated: 5 novices (A), 5 intermediates (B), and 4 experts (C). The results involving intraoperative cholangiography catheter insertion were similar among the 3 groups. As for the other tasks, the expert had better results than the other 2, in which no significant differences occurred. The proposed model is able to discriminate among individuals with different levels of expertise, indicating that the abilities that the model evaluates are relevant in the surgeon's performance in CBD exploration. Construct validity for tasks 2 and 3 was demonstrated. However, task 1 was no capable of distinguishing between groups, and task 4 was not statistically validated.

To assess the construct validity of the URO Mentor (Simbionix Corp., Cleveland, OH, USA) virtual reality training model for several variables of skills training in cysto-urethroscopy, addressing two research questions: (i) Does training on the URO Mentor significantly improve novices' performance in terms of time, trauma, areas inspected and Global Rating Scale (GRS) score?; (ii) is discrimination between different levels of expertise possible using the URO Mentor? Thirty experts and 50 novices performed seven tasks on the URO Mentor during one training session. The first, fourth and seventh tasks were 'test tasks' to evaluate participants' performance. The simulator recorded procedure time and trauma; a supervisor scored which areas were inspected and gave scores on the GRS. A two-way analysis of variance with repeated-measures test was used to analyse experts' and novices' performances, with P < 0.05 considered to indicate statistical significance. Effect sizes (ES) were calculated to quantify the practical significance of the results; ES of 0.10, 0.30, and 0.50 were considered small, medium and large, respectively. Novices' performances showed a significant improvement with large ES in time (linear trend of learning curve P < 0.001, ES 0.66) and mean GRS score (linear trend P < 0.001, ES 0.84, quadratic trend P = 0.018, ES 0.24). There was a medium improvement for trauma (linear trend P < 0.001, ES 0.40) and a small improvement in areas inspected (linear trend P = 0.032, ES 0.21). That the 95% confidence intervals of the measures on the first task of experts and novices did not coincide indicates that differentiation between experts and novices on the four variables measured can be achieved using the URO Mentor. Training on the URO Mentor appears to result in a medium to large improvement of novices' performances for time, trauma, areas inspected and GRS scores. Moreover, discrimination between different levels of expertise is possible using this simulator.

Kiosk 8R-FB-07 - Quantitative Segmentation Analysis in CMR Across Different Levels of Expertise – FBom Beginners to Experts

Bigliani classification is used for determination of acromial morphology, but poor inter-observer reliability has been reported on conventional radiographs. This study aims to assess inter- and intra-observer reliability using magnetic resonance imaging (MRI). Forty consecutive patients diagnosed with subacromial impingement syndrome were included to study. All subjects underwent standard shoulder MRI scan and acromial shape was evaluated by nine observers of different level of expertise (three attending surgeons, three senior orthopaedic residents and three radiologists). A second set of evaluation was performed in order to assess intra-observer reproducibility. Kappa (κ) coefficient analyses both for interobserver reliability and intra-observer reproducibility were then performed. Overall inter-observer agreement among nine observers was fair (κ=0.323). κ values for all 4 individual types ranged from 0.234 to 0.720 with highest agreement for type 4 and lowest agreement for type 3. Second evaluation did not result with an increase of inter-observer agreement (κ=0.338, fair). The κ coefficients for intra-observer reproducibility of nine observers ranged from 0.496 to 0.867. Overall intra-observer reproducibility was substantial. Comparison of inter- and intra-observer reliability among three groups showed no significant difference (p=0.92 and 0.22, respectively). Results showed that MRI did not show superior reliability compared to conventional radiographs. Moreover, inter- and intra-observer agreement did not differ between observers of different level of expertise. Findings of present study suggest that despite a sophisticated imaging modality like MRI, Bigliani's classification apparently lacks accuracy and additional criteria, or different assessment methods are required to assess acromial morphology for clinical guidance.

Icon-function relationship in toolbar icons

The Cancer Genome Atlas (TCGA) community is a joint effort to characterize cancer driver genes in 33 different cancer types from nearly 10,000 exomes [1]. In such an effort, several methods for predicting cancer genes from genomic data are constantly being developed and improved. OpenEBench (https://openebench.bsc.es), is the ELIXIR benchmarking and technical monitoring platform for bioinformatics tools, web servers and workflows. OpenEBench is part of the ELIXIR tools platform ecosystem (https://www.elixir-europe.org/platforms/tools). Its development is led by the INB/ELIXIR-ES (https://inb-elixir.es) Central node at the Barcelona Supercomputing Center (BSC) in collaboration with different partners across Europe. It is focused on providing an infrastructure where end-users can learn from different available options and select the one best fitting their scientific needs; bioinformatics software developers can find relevant datasets and meaningful scientific challenges to evaluate their own developments, and communities interested in a particular scientific domain can easily define which datasets and metrics are relevant for developers to work on, which in turn will allow the field to move ahead. TCGA, as a engaged community in OpenEbench, provides an ideal use-case to demonstrate the latest developments at OpenEBench including the deployment of use of a Virtual Research Environment (VRE). End-users can have access to the software tools who has participated in the TCGA effort to identify cancer driver genes across 33 different cancer types while developers have the opportunity to evaluate their latest developments by submitting their predictions to the OpenEBench VRE. A predefined workflow including format and data validation, metrics computation, results consolidation and visualization is executed for each submission. Results can be visualized at the VRE, downloaded and/or extended by using a number of scripts provided via the OpenEBench repository [2]. In the case of TCGA, users can test their performance against 8 different tools. A list of gene identifiers should be submitted for doing so. They are then compared against several reference datasets defined by the TCGA community depending on the different cancer types, which results in two assessment metrics (true positive rate and precision). This allows the submitted predictions to be benchmarked against the rest of the participants currently present within the TCGA community. The benchmark results are provided in a compressed file (containing the assessment data for all participants and an SVG image with the visualization) and can also be visualized in a scatter-plot where the user can then apply several classification methods implemented in OpenEBench in order to transform the results to table format, which are easier to interpret by non-expert users. [1] Bailey, M. H., Tokheim, C., Porta-Pardo, E., Sengupta, S., Bertrand, D., Weerasinghe, A., … Mariamidze, A. (2018). Comprehensive Characterization of Cancer Driver Genes and Mutations. Cell, 173(2), 371–385.e18. [2] https://github.com/inab/TCGA_visualizer

The purpose of this quantitative research study was to explore instructional computer use by faculty members in the College of Education (COE) at an Anatolian university in Turkey. The following variables were analyzed regarding their influence on the faculty members' use of computers for instructional purposes: demographic characteristics, computer expertise, computer access, barriers to computer access, attitude toward computer use, support for computer use, and adopter categories based on innovativeness. The demographic characteristics of the faculty members included gender, academic rank, department, computer ownership at home and in the office, age, teaching experience, computer experience in general, average number of students taught in one semester, and number of graduate students supervised. The findings from this study show that faculty members in the COE in Turkey reported low levels of use and expertise in instructional computer technologies. The following variables were significantly correlated with the level of computer use by faculty members in the COE: computer ownership in the office, age, years of computer experience in general, computer expertise, computer access, barriers to computer access, attitude toward computer use, support for computer use, and adopter categories based on innovativeness. In this study, group differences also were analyzed for six computer-use factors based on demographic characteristics and adopter categories. Group differences for the computer use factors were found between COE faculty members with regard to academic rank, department, computer ownership at home and in the office, age, teaching experience, computer experience in general, and adopter categories. Associate professors; faculty members who owned a computer in the office or at home; faculty members between the ages

Cognitive science has long shown interest in expertise, in part because prediction and control of expert development would have immense practical value. Most studies in this area investigate expertise by comparing experts with novices. The reliance on contrastive samples in studies of human expertise only yields deep insight into development where differences are important throughout skill acquisition. This reliance may be pernicious where the predictive importance of variables is not constant across levels of expertise. Before the development of sophisticated machine learning tools for data mining larger samples, and indeed, before such samples were available, it was difficult to test the implicit assumption of static variable importance in expertise development. To investigate if this reliance may have imposed critical restrictions on the understanding of complex skill development, we adopted an alternative method, the online acquisition of telemetry data from a common daily activity for many: video gaming. Using measures of cognitive-motor, attentional, and perceptual processing extracted from game data from 3360 Real-Time Strategy players at 7 different levels of expertise, we identified 12 variables relevant to expertise. We show that the static variable importance assumption is false - the predictive importance of these variables shifted as the levels of expertise increased - and, at least in our dataset, that a contrastive approach would have been misleading. The finding that variable importance is not static across levels of expertise suggests that large, diverse datasets of sustained cognitive-motor performance are crucial for an understanding of expertise in real-world contexts. We also identify plausible cognitive markers of expertise.

A quantitative research surveying 390 people with different levels of expertise in computer usage was conducted to understand user behavior from three perspectives: How users make sure that the sites they are using are safe, How users deal with forgotten passwords, How secure is the “security questions”. The finding shows users’ pattern of behavior in checking security when viewing a web application, the way they deal with numerous passwords and retrieval of the forgotten password by using the security question. The research concludes that most people would be able to answer a variety of security questions for other people in their entourage. Users seem to have significantly different behaviors statistically by age group and level of expertise.