Semi-automatic Method Research Articles

Who Does(n't) Target You?

Political campaigns are increasingly investing in targeted advertising on social media platforms to reach voters. Despite critical implications for citizens and elections, little is known about the targeting strategies deployed by political parties — especially in countries beyond the global north. This paper provides a comprehensive descriptive analysis of political microtargeting practices on Facebook and Instagram across 95 countries during 113 national elections. By analyzing the Meta Ad Targeting dataset, we explore targeting strategies of 54k political advertisers who ran 2.5 million ads between August 2020 and December 2022. The findings indicate that election campaigns worldwide utilize targeted advertising. Most commonly, spending is allocated towards a single targeting criterion, however, in wealthier countries and electoral systems with proportional representation, a greater amount of money is spent on microtargeting by combining multiple criteria. Furthermore, targeting strategies vary along ideological lines of political parties who seek out voters more typically aligned with each side of the political spectrum. Nonetheless, parties use microtargeting irrespective of political ideology. Our findings offer the first comparative analysis of political microtargeting on Meta platforms for countries across all continents. Methodologically, we introduce a semi-automatic method to identify worldwide political advertisers using multiple data sources. Our study deepens the understanding of how country and party contexts explain differences in targeting strategies, highlighting the need for more research beyond the global north. Finally, our results have important implications for policy makers, and other stakeholders who seek to develop regulations to address the challenges posed by political microtargeting techniques.

Towards the systematic reconnaissance of seismic signals from glaciers and ice sheets – Part 2: Unsupervised learning for source process characterization

Abstract. Given the high number and diversity of events in a typical cryoseismic dataset, in particular those recorded on ice sheet margins, it is desirable to use a semi-automated method of grouping similar events for reconnaissance and ongoing analysis. We present a workflow for employing semi-unsupervised cluster analysis to inform investigations of the processes occurring in glaciers and ice sheets. In this demonstration study, we make use of a seismic event catalogue previously compiled for the Whillans Ice Stream, for the 2010–2011 austral summer (outlined in Part 1, Latto et al., 2024). We address the challenges of seismic event analysis for a complex wave field by clustering similar seismic events into groups using characteristic temporal, spectral, and polarization attributes of seismic time series with the k-means++ algorithm. This provides the basis for a reconnaissance analysis of a seismic wave field that contains local events (from the ice stream) set in an ambient wave field that itself contains a diversity of signals (mostly from the Ross Ice Shelf). As one result, we find that two clusters include stick-slip events that diverge in terms of length and initiation locality (i.e., central sticky spot and/or the grounding line). We also identify a swarm of high-frequency signals on 16–17 January 2011 that are potentially associated with a surface melt event from the Ross Ice Shelf. Used together with the event detection presented in Part 1, the semi-automated workflow could readily be generalized to other locations and, as a possible benchmark procedure, could enable the monitoring of remote glaciers over time and comparisons between locations.

Manual scoring of periodic limb movements in children: is it still necessary?

ObjectiveAlthough manual scoring has been classically considered the gold standard to identify periodic leg movements (PLM), it is a very time consuming and expensive process, also subject to variability in interpretation. In the last decades, different authors have observed reasonably good agreement between automated PSG scoring algorithms and manual scoring in adults, according to established criteria. We aim to compare the automatic software analysis of our polysomnogram with the manual staging in children with sleep-disordered breathing. MethodsWe performed a semiautomatic method, in which an experienced technician watched the video recording and removed from the automatic analysis those movements that did not correspond to true candidate leg movement (LM). ResultsA total of 131 PSGs were studied; applying the established criteria, 65 children were diagnosed of obstructive sleep apnea, and 66 presented snoring but with no sleep apnea. The mean age was 6.7 years (±1.7) and twenty-five children (19.08 %) had a PLMI >5/h. Statistical differences were found not only for PLMI (manual: 2.20 (0.7, 4.1) vs automatic (6.4 (3.85,9.5); p < 0.001), but for almost of all indexes assessed between the automatic and the manual scoring analysis. The level of concordance was only moderate for PLM index (0.63 [0.51–0.72]); showing that, unlike the articles published in the adult population, automatic analysis is not accurate in children and, manually or semi-automatically analysis as ours need to be done. ConclusionIt seems that PLM detection algorithm might work accurately but, the real need would be a true LM detection algorithm.

A NLP-based semi-automatic identification system for delays in follow-up examinations: an Italian case study on clinical referrals

BackgroundThis study aims to propose a semi-automatic method for monitoring the waiting times of follow-up examinations within the National Health System (NHS) in Italy, which is currently not possible to due the absence of the necessary structured information in the official databases.MethodsA Natural Language Processing (NLP) based pipeline has been developed to extract the waiting time information from the text of referrals for follow-up examinations in the Lombardy Region. A manually annotated dataset of 10 000 referrals has been used to develop the pipeline and another manually annotated dataset of 10 000 referrals has been used to test its performance. Subsequently, the pipeline has been used to analyze all 12 million referrals prescribed in 2021 and performed by May 2022 in the Lombardy Region.ResultsThe NLP-based pipeline exhibited high precision (0.999) and recall (0.973) in identifying waiting time information from referrals’ texts, with high accuracy in normalization (0.948-0.998). The overall reporting of timing indications in referrals’ texts for follow-up examinations was low (2%), showing notable variations across medical disciplines and types of prescribing physicians. Among the referrals reporting waiting times, 16% experienced delays (average delay = 19 days, standard deviation = 34 days), with significant differences observed across medical disciplines and geographical areas.ConclusionsThe use of NLP proved to be a valuable tool for assessing waiting times in follow-up examinations, which are particularly critical for the NHS due to the significant impact of chronic diseases, where follow-up exams are pivotal. Health authorities can exploit this tool to monitor the quality of NHS services and optimize resource allocation.

Semi-automatic Knowledge Graph Construction Based on Deep Learning

The paper studied course knowledge graph in teaching resources and curriculum knowledge management tasks from the perspective of knowledge management. Considering that the course of Python Language Programming itself has formed a relatively complete knowledge system and knowledge point structure, the paper adopted a top-down approach to build the knowledge graph. Firstly, the paper obtained different types of course-related corpus and data from different sources, and then constructed the ontology layer of Python programming course. At the ontology level, the paper defined the concept type, relation type and attribute type of the course domain respectively. Considering the completeness of knowledge points in the curriculum domain knowledge graph, the paper extracted all entities, relationships, attributes, and its values from the curriculum corpus using a semi-automatic extraction method that takes into account both accuracy and efficiency based on the modeling results of the ontology layer. Then they were transformed into triples in the form of &lt; entity, relationship, entity &gt; or &lt; entity, attribute, attribute value &gt; to build data layer of knowledge graph. Finally, visualization of triplet data was realized through Neo4j graph database.

An open-source, three-dimensional growth model of the mandible

The available reference data for the mandible and mandibular growth consists primarily of two-dimensional linear or angular measurements. The aim of this study was to create the first open-source, three-dimensional statistical shape model of the mandible that spans the complete growth period. Computed tomography scans of 678 mandibles from children and young adults between 0 and 22 years old were included in the model. The mandibles were segmented using a semi-automatic or automatic (artificial intelligence-based) segmentation method. Point correspondence among the samples was achieved by rigid registration, followed by non-rigid registration of a symmetrical template onto each sample. The registration process was validated with adequate results. Principal component analysis was used to gain insight in the variation within the dataset and to investigate age-related changes and sexual dimorphism. The presented growth model is accessible globally and free-of-charge for scientists, physicians and forensic investigators for any kind of purpose deemed suitable. The versatility of the model opens up new possibilities in the fields of oral and maxillofacial surgery, forensic sciences or biological anthropology. In clinical settings, the model may aid diagnostic decision-making, treatment planning and treatment evaluation.

Comparison of semi-automatic and manual segmentation methods for tumor delineation on head and neck squamous cell carcinoma (HNSCC) positron emission tomography (PET) images

Objective. Accurate and reproducible tumor delineation on positron emission tomography (PET) images is required to validate predictive and prognostic models based on PET radiomic features. Manual segmentation of tumors is time-consuming whereas semi-automatic methods are easily implementable and inexpensive. This study assessed the reliability of semi-automatic segmentation methods over manual segmentation for tumor delineation in head and neck squamous cell carcinoma (HNSCC) PET images. Approach. We employed manual and six semi-automatic segmentation methods (just enough interaction (JEI), watershed, grow from seeds (GfS), flood filling (FF), 30% SUVmax and 40%SUVmax threshold) using 3D slicer software to extract 128 radiomic features from FDG-PET images of 100 HNSCC patients independently by three operators. We assessed the distributional properties of all features and considered 92 log-transformed features for subsequent analysis. For each paired comparison of a feature, we fitted a separate linear mixed effect model using the method (two levels; manual versus one semi-automatic method) as a fixed effect and the subject and the operator as the random effects. We estimated different statistics—the intraclass correlation coefficient agreement (aICC), limits of agreement (LoA), total deviation index (TDI), coverage probability (CP) and coefficient of individual agreement (CIA)—to evaluate the agreement between the manual and semi-automatic methods. Main results. Accounting for all statistics across 92 features, the JEI method consistently demonstrated acceptable agreement with the manual method, with median values of aICC = 0.86, TDI = 0.94, CP = 0.66, and CIA = 0.91. Significance. This study demonstrated that JEI method is a reliable semi-automatic method for tumor delineation on HNSCC PET images.

A comparison of semi-automatic and manual analysis of right ventricular strain in speckle tracking echocardiography

The aim was to compare the reproducibility of the traditional manual analysis of the right ventricle (RV) strain with a semi-automatic analysis (an automatic method with the possibility of adjustment) in speckle-tracking echocardiography (STE).Materials and methods. Evaluation of RV by speckle-tracking echocardiography was performed in 71 patients with two approaches: manual (Q-Analysis) and semi-automatic (AutoStrainRV). Such parameters as RV global longitudinal strain (RVGLS), RV longitudinal free wall strain (RVLFWS), RV longitudinal free wall segmental strain, and the amplitude of movement of the tricuspid annular plane systolic excursion (TAPSE) were compared.Results. There were no significant differences between the values of RVGLS (21.9 ± 4.7% vs. 21.6 ± 3.6%, p = 0.488, respectively) and RVLFWS (24.6 ± 5.1% vs. 25.3 ± 4.6%, p = 0.212, respectively) obtained by manual and semi-automatic strain analysis. The significant differences were revealed in the values of RVLFWS in the basal and middle segments estimated by a semi-automatic method and by a manual method (22.6 ± 7.1% vs. 27.0 ± 10.1%, p &lt; 0.002, and 23.8 ± 5.9% vs. 25.8 ± 8.0%, p &lt; 0.05, respectively). Bland-Altman analysis showed mean bias for RVGLS -0.31 (95% CI: -7.62–7.00) and for RVLFWS - 0.70 (95% CI: -8.44–9.83). The values of the TAPSE obtained by semi-automatic analysis were lower and significantly different from the values calculated in the M-mode (22.6 ± 3.8 mm vs. 24.9 ± 4.5 mm, p &lt; 0.001). Strain analysis by the semi-automatic method requires less time than the manual method (22.9 ± 4.5 seconds vs. 38.9 ± 7.8 seconds, p &lt; 0.05).Conclusions. The choice of approach for RV strain assessment (manual or semi-automatic) does not significantly affect the values of RV strain but may affect the result of segmental RV strain in the basal and middle segments. The semi-automatic strain analysis of RV provides a faster and more reproducible assessment of RV function. The values of the TAPSE by semi-automatic strain analysis are significantly lower compared to the values of the TAPSE in M-mode.

Vein of Galen aneurysmal malformation: does size affect outcome?

PurposeTo validate a semiautomated method for segmenting vein of Galen aneurysmal malformations (VGAM) and to assess the relationship between VGAM volume and other angioarchitectural features, cardiological findings, and outcomes.MethodsIn this retrospective study, we selected all subjects with VGAM admitted to the Gaslini Children’s Hospital between 2009 and 2022. Clinical data were retrieved from electronic charts. We compared 3D-Slicer segmented VGAM volumes obtained by two independent observers using phase-contrast MR venography to those obtained with manual measurements performed on T2-weighted images. The relationship between VGAM volumes and clinical and neuroimaging features was then explored.ResultsForty-three subjects with VGAM (22 males, mean age 6.56 days) were included in the study. Manual and semiautomated VGAM volumes were well correlated for both readers (r = 0.86 and 0.82, respectively). Regarding reproducibility, the inter-rater interclass correlation coefficients were 0.885 for the manual method and 0.992 for the semiautomated method (p < 0.001). The standard error for repeated measures was lower for the semiautomated method (0.04 versus 0.40 of manual method). Higher VGAM volume was associated with superior sagittal sinus narrowing, jugular bulb stenosis, and aqueductal stenosis (p < 0.05). A weak correlation was found between VGAM volume and straight sinus dilatation (r = 0.331) and superior sagittal sinus index (r = − 0.325). No significant associations were found with cardiac findings, post-embolization complications, and outcome (p > 0.05).ConclusionsSemiautomated VGAM volumetry is feasible and reliable with improved reproducibility compared to the manual method. VGAM volume is not a prognostic factor for clinical outcome, but it is related to other venous findings with potential hemodynamic effects.

Development of a semi-automated tool to measure fetal abdominal wall thickness during ultrasound at 20 weeks' gestation.

To develop a semi-automated tool for measuring fetal abdominal wall thickness (AWT). To validate the software using images captured by other centers and create a nomogram for fetal AWT between 18 and 20 weeks. A semiautomated tool that measured AWT was developed using images captured at the routine 20-week morphology scan. The software was developed using digital images captured routinely during scans of low-risk women. Inter- and intraobserver reliability was assessed between manual and semi-automated measures. The tool was validated using images acquired from other centers. Linear regression and quadratic polynomials were used to create a nomogram for AWT. The semi-automated tool was able to measure AWT in all images. Interoperator reliability was 0.90 and 0.97 (P < 0.05) for manual and semi-automated methods, respectively. Measurement agreement varied between three operators from moderate to excellent (0.77, 0.87, 0.92), with overall agreement being good (0.85). The tool could be successfully applied to 89% of images from other centers. A nomogram was generated for AWT measurements of fetuses at 18-20 weeks in normal, low risk mothers. Semi-automated measurement of AWT was feasible using images captured during the routine 20-week scan. This approach had lower inter- and intraobserver variability compared to manual measurement.

PyOKR: A Semi-Automated Method for Quantifying Optokinetic Reflex Tracking Ability.

The study of behavioral responses to visual stimuli is a key component of understanding visual system function. One notable response is the optokinetic reflex (OKR), a highly conserved innate behavior necessary for image stabilization on the retina. The OKR provides a robust readout of image tracking ability and has been extensively studied to understand visual system circuitry and function in animals from different genetic backgrounds. The OKR consists of two phases: a slow tracking phase as the eye follows a stimulus to the edge of the visual plane and a compensatory fast phase saccade that resets the position of the eye in the orbit. Previous methods of tracking gain quantification, although reliable, are labor intensive and can be subjective or arbitrarily derived. To obtain more rapid and reproducible quantification of eye tracking ability, we have developed a novelsemi-automated analysis program, PyOKR, that allows for quantification of two-dimensional eye tracking motion in response to any directional stimulus, in addition to being adaptable to any type of video-oculography equipment. This method provides automated filtering, selection of slow tracking phases, modeling of vertical and horizontal eye vectors, quantification of eye movement gains relative to stimulus speed, and organization of resultant data into a usable spreadsheet for statistical and graphical comparisons. This quantitative and streamlined analysis pipeline, readily accessible via PyPI import, provides a fast and direct measurement of OKR responses, thereby facilitating the study of visual behavioral responses.

Automated Growth Rate Measurement of the Facet Surfaces of Single Crystals of the β-Form of l-Glutamic Acid Using Machine Learning Image Processing.

Precision measurement of the growth rate of individual single crystal facets (hkl) represents an important component in the design of industrial crystallization processes. Current approaches for crystal growth measurement using optical microscopy are labor intensive and prone to error. An automated process using state-of-the-art computer vision and machine learning to segment and measure the crystal images is presented. The accuracies and efficiencies of the new crystal sizing approach are evaluated against existing manual and semi-automatic methods, demonstrating equivalent accuracy but over a much shorter time, thereby enabling a more complete kinematic analysis of the overall crystallization process. This is applied to measure in situ the crystal growth rates and through this determining the associated kinetic mechanisms for the crystallization of β-form l-glutamic acid from the solution phase. Growth on the {101} capping faces is consistent with a Birth and Spread mechanism, in agreement with the literature, while the growth rate of the {021} prismatic faces, previously not available in the literature, is consistent with a Burton-Cabrera-Frank screw dislocation mechanism. At a typical supersaturation of σ = 0.78, the growth rate of the {101} capping faces (3.2 × 10-8 m s-1) is found to be 17 times that of the {021} prismatic faces (1.9 × 10-9 m s-1). Both capping and prismatic faces are found to have dead zones in their growth kinetic profiles, with the capping faces (σc = 0.23) being about half that of the prismatic faces (σc = 0.46). The importance of this overall approach as an integral component of the digital design of industrial crystallization processes is highlighted.

Meso-scale seabed quantification with geoacoustic inversion

Knowledge of sub-seabed geoacoustic properties, for example depth dependent sound speed and porosity, is of importance for a variety of applications. Here, we present a semi-automated geoacoustic inversion method for autonomous underwater vehicle data that objectively adapts model inference to seabed structure. Through parallelized trans-dimensional Bayesian inference, we infer seabed properties along a 12 km survey track on the scale of about 10 cm and 50 m in the vertical and horizontal, respectively. The inferred seabed properties include sound speed, attenuation, density, and porosity as a function of depth from acoustic reflection coefficient data. Parameter uncertainties are quantified, and the seabed properties agree closely with core samples at two control points and the layering structure with an independent sub-bottom seismic survey. Recovering high resolution seabed properties over large areas is shown to be feasible, which could become an important tool for marine industries, navies and oceanic research organizations.

Myosteatosis and aortic calcium score on abdominal CT as prognostic markers in non-dialysis chronic kidney disease patients

We aimed to examine the relationship between abdominal computed tomography (CT)-based body composition data and both renal function decline and all-cause mortality in patients with non-dialysis chronic kidney disease (CKD). This retrospective study comprised non-dialysis CKD patients who underwent consecutive unenhanced abdominal CT between January 2010 and December 2011. CT-based body composition was measured using semiautomated method that included visceral fat, subcutaneous fat, skeletal muscle area and density, and abdominal aortic calcium score (AAS). Sarcopenia and myosteatosis were defined by decreased skeletal muscle index (SMI) and decreased skeletal muscle density, respectively, each with specific cutoffs. Risk factors for CKD progression and survival were identified using logistic regression and Cox proportional hazard regression models. Survival between groups based on myosteatosis and AAS was compared using the Kaplan–Meier curve. 149 patients (median age: 70 years) were included; 79 (53.0%) patients had sarcopenia and 112 (75.2%) had myosteatosis. The median AAS was 560.9 (interquartile range: 55.7–1478.3)/m2. The prognostic factors for CKD progression were myosteatosis [odds ratio (OR) = 4.31, p = 0.013] and high AAS (OR = 1.03, p = 0.001). Skeletal muscle density [hazard ratio (HR) = 0.93, p = 0.004] or myosteatosis (HR = 4.87, p = 0.032) and high AAS (HR = 1.02, p = 0.001) were independent factors for poor survival outcomes. The presence of myosteatosis and the high burden of aortic calcium were significant factors for CKD progression and survival in patients with non-dialysis CKD.

A Growth Prediction Model of Pulmonary Ground-Glass Nodules Based on Clinical Visualization Parameters

Objective To establish a model for predicting the growth of pulmonary ground-glass nodules (GGN) based on the clinical visualization parameters extracted by the 3D reconstruction technique and to verify the prediction performance of the model. Methods A retrospective analysis was carried out for 354 cases of pulmonary GGN followed up regularly in the outpatient of pulmonary nodules in Zhoushan Hospital of Zhejiang Province from March 2015 to December 2022.The semi-automatic segmentation method of 3D Slicer was employed to extract the quantitative imaging features of nodules.According to the follow-up results,the nodules were classified into a resting group and a growing group.Furthermore,the nodules were classified into a training set and a test set by the simple random method at a ratio of 7∶3.Clinical and imaging parameters were used to establish a prediction model,and the prediction performance of the model was tested on the validation set. Results A total of 119 males and 235 females were included,with a median age of 55.0 (47.0,63.0) years and the mean follow-up of (48.4±16.3) months.There were 247 cases in the training set and 107 cases in the test set.The binary Logistic regression analysis showed that age (95%CI=1.010-1.092,P=0.015) and mass (95%CI=1.002-1.067,P=0.035) were independent predictors of nodular growth.The mass (M) of nodules was calculated according to the formula M=V×(CTmean+1000)×0.001 (where V is the volume,V=3/4πR3,R:radius).Therefore,the logit prediction model was established as ln[P/(1-P)]=-1.300+0.043×age+0.257×two-dimensional diameter+0.007×CTmean.The Hosmer-Lemeshow goodness of fit test was performed to test the fitting degree of the model for the measured data in the validation set (χ2=4.515,P=0.808).The check plot was established for the prediction model,which showed the area under receiver-operating characteristic curve being 0.702. Conclusions The results of this study indicate that patient age and nodule mass are independent risk factors for promoting the growth of pulmonary GGN.A model for predicting the growth possibility of GGN is established and evaluated,which provides a basis for the formulation of GGN management strategies.

Estimation of EIA by Semi-Automated Direct Method for Western Indian Urban Catchment Area

Urbanization impacts water resources, with impervious surfaces indicating the degree of disturbance. Effective impervious area (EIA) is a better measure than total impervious area (TIA) for urban runoff. EIA is the portion of TIA connected to the storm sewer system. Remote sensing data alone is insufficient for EIA determination. An alternative approach integrates remote sensing data with drainage networks to ensure connectivity of impervious surfaces. This approach uses GIS data including multi-sensor data. The modified literature algorithm was used in python to analyze the study area. It is observed that for western Indian urban catchment area, some empirical equation such as Sutherland’s and Alley-Veenhuis showed under-estimation and over-estimation at some watershed parts which were obtained using DEM delineation. Based on the results obtained from proposed method, a quadratic relationship was established between easily measured TIA and hydraulically relevant EIA for an urban catchment in western India. These outcomes should result in more effective and properly designed SCMs, with potential improvements in water quality and cost savings.

Development and validation of a semi-automated and unsupervised method for femur segmentation from CT

Quantitative computed tomography (QCT)-based in silico models have demonstrated improved accuracy in predicting hip fractures with respect to the current gold standard, the areal bone mineral density. These models require that the femur bone is segmented as a first step. This task can be challenging, and in fact, it is often almost fully manual, which is time-consuming, operator-dependent, and hard to reproduce. This work proposes a semi-automated procedure for femur bone segmentation from CT images. The proposed procedure is based on the bone and joint enhancement filter and graph-cut algorithms. The semi-automated procedure performances were assessed on 10 subjects through comparison with the standard manual segmentation. Metrics based on the femur geometries and the risk of fracture assessed in silico resulting from the two segmentation procedures were considered. The average Hausdorff distance (0.03 ± 0.01 mm) and the difference union ratio (0.06 ± 0.02) metrics computed between the manual and semi-automated segmentations were significantly higher than those computed within the manual segmentations (0.01 ± 0.01 mm and 0.03 ± 0.02). Besides, a blind qualitative evaluation revealed that the semi-automated procedure was significantly superior (p < 0.001) to the manual one in terms of fidelity to the CT. As for the hip fracture risk assessed in silico starting from both segmentations, no significant difference emerged between the two (R2 = 0.99). The proposed semi-automated segmentation procedure overcomes the manual one, shortening the segmentation time and providing a better segmentation. The method could be employed within CT-based in silico methodologies and to segment large volumes of images to train and test fully automated and supervised segmentation methods.

A semi-automatic method for extracting mitochondrial cristae characteristics from 3D focused ion beam scanning electron microscopy data

Mitochondria are the main suppliers of energy for cells and their bioenergetic function is regulated by mitochondrial dynamics: the constant changes in mitochondria size, shape, and cristae morphology to secure cell homeostasis. Although changes in mitochondrial function are implicated in a wide range of diseases, our understanding is challenged by a lack of reliable ways to extract spatial features from the cristae, the detailed visualization of which requires electron microscopy (EM). Here, we present a semi-automatic method for the segmentation, 3D reconstruction, and shape analysis of mitochondria, cristae, and intracristal spaces based on 2D EM images of the murine hippocampus. We show that our method provides a more accurate characterization of mitochondrial ultrastructure in 3D than common 2D approaches and propose an operational index of mitochondria’s internal organization. With an improved consistency of 3D shape analysis and a decrease in the workload needed for large-scale analysis, we speculate that this tool will help increase our understanding of mitochondrial dynamics in health and disease.

NOVEL ECHOCARDIOGRAM ANALYSIS OF CARDIAC DYSFUNCTION IS ASSOCIATED WITH MORTALITY IN PEDIATRIC SEPSIS.

Objectives: The objective of our study was to semiautomatically generate echocardiogram indices in pediatric sepsis using novel algorithms and determine which indices were associated with mortality. We hypothesized that strain and diastolic indices would be most associated with mortality. Design: Retrospective cohort study of children with sepsis from 2017 to 2022. Survivors and nonsurvivors were compared for echocardiogram indices. Multivariate Cox proportional hazard models were constructed for our primary outcome of in-hospital mortality. Linear regression was performed for secondary outcomes, which included multiple composite 28-day outcomes. Results: Of the 54 patients in the study, 9 (17%) died. Multiple echocardiogram indices of both right (RV) and left ventricles (LV) were associated with in-hospital mortality [RV GLS adjusted hazard ratio (aHR): 1.16 (1.03-1.29), P = 0.011; RV global longitudinal early diastolic strain rate (GLSre) aHR: 0.24 (0.07 to 0.75), P = 0.014; LV GLSre aHR: 0.33 (0.11-0.97), P = 0.044]. Impairment in GLS was associated with fewer ventilator-free days [RV GLS β-coefficient: -0.47 (-0.84 to -0.10), P = 0.013; LV GLS β-coefficient -0.62 (-1.07 to -0.17), P = 0.008], organ-support free days [RV GLS β-coefficient: -0.49 (-0.87 to -0.11), P = 0.013; LV GLS β-coefficient: -0.64 (-1.10 to -0.17), P = 0.008], and days free from ICU [RV GLS β-coefficient: -0.42 (-0.79 to -0.05), P = 0.026; LV GLS β-coefficient: -0.58 (-1.03 to -0.13), P = 0.012]. Systolic indices were not associated with mortality in this cohort. Conclusion: Our study demonstrates the feasibility of obtaining echocardiogram indices in a semiautomatic method using our algorithms. We showed that abnormal strain is associated with worse outcomes in a cohort of children with sepsis.

NaRuto: Automatically Acquiring Planning Models from Narrative Texts

Domain model acquisition has been identified as a bottleneck in the application of planning technology, especially within narrative planning. Learning action models from narrative texts in an automated way is essential to overcome this barrier, but challenging because of the inherent complexities of such texts. We present an evaluation of planning domain models derived from narrative texts using our fully automated, unsupervised system, NaRuto. Our system combines structured event extraction, predictions of commonsense event relations, and textual contradictions and similarities. Evaluation results show that NaRuto generates domain models of significantly better quality than existing fully automated methods, and even sometimes on par with those created by semi-automated methods, with human assistance.