Three-Dimensional Reconstruction Pre-Training as a Prior to Improve Robustness to Adversarial Attacks and Spurious Correlation.

Ensuring robustness of image classifiers against adversarial attacks and spurious correlation has been challenging. One of the most effective methods for adversarial robustness is a type of data augmentation that uses adversarial examples during training. Here, inspired by computational models of human vision, we explore a synthesis of this approach by leveraging a structured prior over image formation: the 3D geometry of objects and how it projects to images. We combine adversarial training with a weight initialization that implicitly encodes such a prior about 3D objects via 3D reconstruction pre-training. We evaluate our approach using two different datasets and compare it to alternative pre-training protocols that do not encode a prior about 3D shape. To systematically explore the effect of 3D pre-training, we introduce a novel dataset called Geon3D, which consists of simple shapes that nevertheless capture variation in multiple distinct dimensions of geometry. We find that while 3D reconstruction pre-training does not improve robustness for the simplest dataset setting, we consider (Geon3D on a clean background) that it improves upon adversarial training in more realistic (Geon3D with textured background and ShapeNet) conditions. We also find that 3D pre-training coupled with adversarial training improves the robustness to spurious correlations between shape and background textures. Furthermore, we show that the benefit of using 3D-based pre-training outperforms 2D-based pre-training on ShapeNet. We hope that these results encourage further investigation of the benefits of structured, 3D-based models of vision for adversarial robustness.

Is Mean Platelet Volume a Predictive Marker for the Development of Thrombosis in Patients with COVID-19 Infection?

Mean platelet volume (MPV) can provide important information about the course and prognosis of many diseases. MPV is an early indicator of platelet activation, which has an important role in the pathogenesis of thrombosis. In this study, we aimed to investigate whether MPV was a predictive marker for the development of thrombosis in hospitalized patients with COVID-19 infection. Fifty-seven patients whose courses were followed after the diagnosis of COVID-19 infection using a polymerase chain reaction test during the pandemic were included in the study. Our results demonstrated that there was a negative correlation between platelet count and MPV (r=0.470, p≤ 0.01), and there was a positive correlation between Platelet Distribution Width (PDW) and MPV (r=0,933, p≤ 0.01), but no significant correlation was found between the other variables and MPV.

From task structures to world models: what do LLMs know?

In what sense does a large language model (LLM) have knowledge? We answer by granting LLMs 'instrumental knowledge': knowledge gained by using next-word generation as an instrument. We then ask how instrumental knowledge is related to the ordinary, 'worldly knowledge' exhibited by humans, and explore this question in terms of the degree to which instrumental knowledge can be said to incorporate the structured world models of cognitive science. We discuss ways LLMs could recover degrees of worldly knowledge and suggest that such recovery will be governed by an implicit, resource-rational tradeoff between world models and tasks. Our answer to this question extends beyond the capabilities of a particular AI system and challenges assumptions about the nature of knowledge and intelligence.

Impact of lipodystrophy on health-related quality of life: the QuaLip study.

BACKGROUND: Lipodystrophy is a rare disease characterized by loss of adipose tissue. Natural history studies have demonstrated significant burden of disease; however, there is limited data on the impact of lipodystrophy on quality of life (QoL) and psychoemotional well-being. The QuaLip study is a prospective observational real-world study that aims to determine the impact of lipodystrophy on QoL and psychoemotional well-being and explore subjective burden of the disease. Sixty-seven adult patients and eight pediatric patients with lipodystrophy were included. Patients were followed up for 24 months and assessments were repeated every three months. Patients were examined by licensed psychiatrists at baseline, and at year 1 and year 2 visits. RESULTS: Eighteen (27.69%) of 65 adult patients (two subjects refused psychiatric assessment) were diagnosed with a psychiatric disorder (e.g., depressive episodes, mixed anxiety and depressive disorder, anxiety disorder, adjustment disorder, recurrent depression, panic disorder, generalized anxiety disorder, unspecified mood disorder, nonorganic sleep disorder, post-traumatic stress disorder, depressive episode comorbidity, social phobia and obsessive-compulsive disorder comorbidity). Lipodystrophy disease and QoL questionnaires revealed a significant disease burden over the study period. More than one-third of patients reported depression symptoms on the Beck Depression Inventory and more than one-fourth of the patients reported significant hunger throughout the study period. Physical appearance, fatigue, and pain contributed to the disease burden. QoL scores were lower in patients with psychiatric disease and in those with poor metabolic control. Attention deficit hyperactivity disorder, depressive disorder, sub-threshold depressive symptoms, obsessive-compulsive disorder, appetite problems, and issues with physical appearance were identified in selected pediatric subjects. CONCLUSIONS: Lipodystrophy has a significant impact on QoL and psychoemotional well-being. Psychiatric disorders seem to be underdiagnosed among patients with lipodystrophy.

How does the primate brain combine generative and discriminative computations in vision?

Vision is widely understood as an inference problem. However, two contrasting conceptions of the inference process have each been influential in research on biological vision as well as the engineering of machine vision. The first emphasizes bottom-up signal flow, describing vision as a largely feedforward, discriminative inference process that filters and transforms the visual information to remove irrelevant variation and represent behaviorally relevant information in a format suitable for downstream functions of cognition and behavioral control. In this conception, vision is driven by the sensory data, and perception is direct because the processing proceeds from the data to the latent variables of interest. The notion of "inference" in this conception is that of the engineering literature on neural networks, where feedforward convolutional neural networks processing images are said to perform inference. The alternative conception is that of vision as an inference process in Helmholtz's sense, where the sensory evidence is evaluated in the context of a generative model of the causal processes that give rise to it. In this conception, vision inverts a generative model through an interrogation of the sensory evidence in a process often thought to involve top-down predictions of sensory data to evaluate the likelihood of alternative hypotheses. The authors include scientists rooted in roughly equal numbers in each of the conceptions and motivated to overcome what might be a false dichotomy between them and engage the other perspective in the realm of theory and experiment. The primate brain employs an unknown algorithm that may combine the advantages of both conceptions. We explain and clarify the terminology, review the key empirical evidence, and propose an empirical research program that transcends the dichotomy and sets the stage for revealing the mysterious hybrid algorithm of primate vision.

Perception of 3D shape integrates intuitive physics and analysis-by-synthesis.

Many surface cues support three-dimensional shape perception, but humans can sometimes still see shape when these features are missing-such as when an object is covered with a draped cloth. Here we propose a framework for three-dimensional shape perception that explains perception in both typical and atypical cases as analysis-by-synthesis, or inference in a generative model of image formation. The model integrates intuitive physics to explain how shape can be inferred from the deformations it causes to other objects, as in cloth draping. Behavioural and computational studies comparing this account with several alternatives show that it best matches human observers (total n = 174) in both accuracy and response times, and is the only model that correlates significantly with human performance on difficult discriminations. We suggest that bottom-up deep neural network models are not fully adequate accounts of human shape perception, and point to how machine vision systems might achieve more human-like robustness.

The pH of Drinking Water and Its Effect on the pH of Urine.

AIM: The aim was to determine whether urine pH changed or not with different pH values of drinking water. With the results obtained from animal studies, comments can be made about the effect of water with different pH levels that people drink on kidney stones. METHOD: A total of 24 Wistar Albino rats were divided into three groups containing eight rats each: the first group was given water with pH 5.5, the second group was given water with pH 7 and the third group was given water with pH 8.2 in the same environment and conditions during 13 days. All rats consumed water in line with their natural feeding habits. All rats had urine pH measurements performed and recorded every day at the same time. The groups were later compared in terms of daily pH values. RESULTS: When daily urine pH values were compared, there were statistically significant differences between pH measurements on the first, fourth and seventh day (p=0.02, p=0.017 and p=0.007, respectively). When first-day values are compared with post-hoc analyses, the urine pH in Group 2 was identified to be lower compared to Group 1 and Group 3 (p<0.001). When the fourth-day values were assessed, the urine pH of Group 2 was observed to be higher than Group 1 and Group 3 (p<0.001). On the seventh day, Group 3 had higher urine pH compared to the other groups (p<0.001). CONCLUSION: The variation in drinking water pH does not directly change urine pH; however, it causes a change in the urine pH on different days.

Patients with severe coronavirus disease 2019 have high frequency of factor 5 Leiden and prothrombin gene mutations.

We investigated the frequency of factor 5 Leiden (FVL) and prothrombin gene (PTG) mutations in patients with severe coronavirus disease 2019 (COVID-19). Our primary aim is to reveal whether these mutations are associated with severity of disease and mortality. A total of 249 patients were included in this cross-sectional study. Severe COVID-19 cases (with oxygen saturation of less than 90 mmHg and who received ventilation support invasively or noninvasively) were included. FVL and PTG mutations were identified by real time- PCR technique. Frequency of mutations for FVL was 11.7%, whereas for PTG was 3.5%. The frequency of FVL and PTG's mutations in our patient group was found to be significantly higher than the normal population ( P  < 0.0001, 0.004, respectively). There was no difference in the frequency of mutations of FVL and PTG between the patients ventilated - invasively and noninvasively. There was also no difference in D-dimer, ferritin, fibrinogen, ex status, and entubational status between the groups of FVL and PTG mutated and wild-type. To the best of our knowledge, it is the first time that we have examined the frequencies of FVL and PGM's mutations in severe COVID-19 disease on such a large scale. The frequencies of both mutations in severe COVID-19 patients were higher than in the healthy population. We believe that studies prospectively designed, including asymptomatic and mild COVID-19 patients, will provide more comprehensive information on the subject.

Seeing Soft Materials Draped Over Objects: A Case Study of Intuitive Physics in Perception, Attention, and Memory.

We typically think of intuitive physics in terms of high-level cognition, but might aspects of physics also be extracted during lower-level visual processing? Might we not only think about physics, but also see it? We explored this using multiple tasks in online adult samples with objects covered by soft materials-as when you see a chair with a blanket draped over it-where you must account for the physical interactions between cloth, gravity, and object. In multiple change-detection experiments (n = 200), observers from an online testing marketplace were better at detecting image changes involving underlying object structure versus those involving only the superficial folds of cloths-even when the latter were more extreme along several dimensions. And in probe-comparison experiments (n = 100), performance was worse when both probes (vs. only one) appeared on image regions reflective of underlying object structure (equating visual properties). This work collectively shows how vision uses intuitive physics to recover the deeper underlying structure of scenes.

A 5-year surveillance of healthcare-associated infections in a university hospital: A retrospective analysis.

Objectives: "Nosocomial infections" or "healthcare-associated infections" are a significant public health problem around the world. This study aimed to assess the rate of laboratory-confirmed healthcare-associated infections, frequency of nosocomial pathogens, and the antimicrobial resistance patterns of bacterial isolates in a University Hospital. Methods: A retrospective evaluation of healthcare-associated infections in a University Hospital, between the years 2015 and 2019 in Tekirdag, Turkey. Results: During the 5 years, the incidence densities of healthcare-associated infections in intensive care units and clinics were 10.31 and 1.70/1000 patient-days, respectively. The rates of ventilator-associated pneumonia, central line-associated bloodstream infections, and catheter-associated urinary tract infections in intensive care units were 11.57, 4.02, and 1.99 per 1000 device-days, respectively. The most common healthcare-associated infections according to the primary sites were bloodstream infections (55.3%) and pneumonia (20.4%). 67.5% of the isolated microorganisms as nosocomial agents were Gram-negative bacteria, 24.9% of Gram-positive bacteria, and 7.6% of Candida. The most frequently isolated causative agents were Escherichia coli (16.7%) and Pseudomonas aeruginosa (15.7%). The rate of extended-spectrum beta-lactamase production among E. coli isolates was 51.1%. Carbapenem resistance was 29.8% among isolates of P. aeruginosa, 95.1% among isolates of Acinetobacter baumannii, and 18.2% among isolates of Klebsiella pneumoniae. Colistin resistance was 2.4% among isolates of A. baumannii. Vancomycin resistance was 5.3% among isolates of Enterococci. Conclusion: Our study results demonstrate that healthcare-associated infections are predominantly originated by intensive care units. The microorganisms isolated from intensive care units are highly resistant to many antimicrobial agents. The rising incidence of multidrug-resistant microorganisms indicates that more interventions are urgently needed to reduce healthcare-associated infections in our intensive care units.

The incidence of acute kidney injury and its association with mortality in patients diagnosed with COVID-19 followed up in intensive care unit.

INTRODUCTION: The kidneys are some of the most frequently affected organs during coronavirus disease 2019 (COVID-19). This multicenter study evaluated the incidence of and risk factors for acute kidney injury (AKI) in COVID-19 patients followed up in intensive care unit (ICU) and its association with mortality. METHODS: Three hundred twenty-eight patients diagnosed with COVID-19 and hospitalized in ICU were included. Risk factors associated with AKI and mortality were evaluated. RESULTS: Eighty-eight patients (27.9%) were diagnosed with AKI. AKI was significantly associated with older age, higher baseline creatinine level, lower albumin level, and coexistence of cardiovascular disease and chronic obstructive pulmonary disease. Mortality in the entire study group was significantly associated with AKI, older age, requirement of invasive mechanical ventilation, higher neutrophil level, lower lymphocyte, and albumin levels. CONCLUSION: AKI is frequently seen during the course of COVID-19 and is associated with high mortality. Identifying AKI-related risk factors appears essential in the management of COVID-19 patients.

The place of infectious markers in predicting culture positivity in patients with renal failure hospitalized with suspected infection.

Infectious diseases are an important cause of mortality in patients with renal failure. The markers used to diagnose infection in patients with renal failure have various limitations. Culture positivity is an objective guide in that context. The purpose of this study was to examine the effectiveness of frequently used markers of bacterial infection in predicting culture positivity in renal failure patients with renal failure hospitalized with suspected bacterial infection over an approximately 1.5-year period were included in this prospective observational study. Patients' demographic and laboratory findings were recorded. Demographic and laboratory findings and mortality were compared between patients with and without culture-positivity. Parameters affecting culture positivity were also analyzed. Four hundred twenty-six patients (median age 67.50, 45.5% female) were included in the study. Culture positivity was determined in 54.5% of patients. Hospital stay was longer (p < 0.001) and leukocyte (p < 0.001), neutrophil percentage (p < 0.05) and CRP (p < 0.001) values were significantly higher in culture-positive patients. Mortality was also significantly higher in culture-positive patients than in culture-negative patients (p < 0.05). CRP was determined as a predictor of culture positivity at logistic regression analysis (p = 0.000, exp ß [1.004]). Culture positivity was determined in more than half of the patients hospitalized with suspected bacterial infection. CRP, a longstanding marker, was identified as a parameter predicting culture positivity. We think that the determination in further studies of a cut-off point for CRP in determining culture positivity may be a useful diagnostic guide.

Relationship between serum soluble endothelial protein C receptor level and COVID-19 findings.

Coronavirus-related disease-2019 (COVID-19)-associated coagulopathy presents predominantly with thrombosis and leads to complications in close association with inflammatory process. Soluble endothelial protein C receptor (sEPCR), which is the soluble form of EPCR, reduces the anticoagulant and anti-inflammatory activity of activated protein C. The purpose of this study is to investigate the relationship between sEPCR and the laboratory parameters and thorax computed tomography (CT) findings in the course of COVID-19. Twenty-five laboratory-confirmed [reverse transcription-quantitative polimerase chain reaction (RT-qPCR) positive] and 24 clinically diagnosed (RT-qPCR negative) COVID-19 patients were enrolled in the study. Blood specimens were collected for sEPCR and haematological and biochemical parameter measurement. Thorax CT was performed to detect COVID-19 findings. These parameters from RT-qPCR positive and negative patients were then compared. Although there was no difference between the groups in terms of symptoms, the time between the onset of symptoms and the admission time was shorter in RT-qPCR positive group (P = 0.000). sEPCR levels were significantly higher in the RT-qPCR positive group (P = 0.011). Patients with ground-glass opacity and bilateral involvement on thorax CT have higher serum sEPCR levels (P = 0.012 and 0.043, respectively). This study has shown for the first time that serum sEPCR levels, which is a member of coagulation cascade and has also been reported to be associated with inflammation, is higher in patients with positive RT-qPCR test and patients with GGO or bilateral involvement on thorax CT regardless of the PCR result.

Analysis of the coracoid morphology with multiplanar 2D CT and its effects on the graft size in the Latarjet procedure.

PURPOSE: In this study, we aimed to reveal the individual differences regarding the size of the coracoid and their effects on the classical and modified Latarjet procedures. METHODS: Computed tomography images of 120 patients (mean age: 41.18 ± 12.01 years) without shoulder complaints or shoulder instability were evaluated retrospectively. The glenoid width, the surgical graft length, and the coracoid total length, width, and thickness were measured using the multiplanar reconstruction method on the Sectra Picture Archiving and Communications System (PACS) system. Age, gender, side, the dominant hand, and the height of the patients were recorded and the correlations between them were investigated. On the created hypothetical model, the current size of the coracoid was evaluated to determine what size of glenoid defects it could repair by employing the classical and the modified Latarjet techniques. RESULTS: There was no significant difference between the right-hand-dominant group and the left-hand-dominant group in terms of coracoid measurement results (p > 0.05). Again, there was no statistically significant difference between the right and the left side regarding the coracoid size (p > 0.05). A positive correlation could be detected only between age and the coracoid width and thickness (p < 0.05). A positive correlation was also found between the glenoid width and the coracoid width and thickness in both shoulders (p < 0.001). Coracoid thickness could fill in the defects that amounted to 40% of the glenoid width, while the coracoid width could fill in for the defects that were 50% of the glenoid width in both genders. CONCLUSION: Our study showed that hand dominance and side were not effective on the coracoid dimensions. In addition, it has been shown that the coracoid dimensions did not have a significant effect in the choice of Latarjet technique in terms of defect repair and that repair rates of up to 40% could be achieved in glenoid defects with both techniques.

Efficient inverse graphics in biological face processing.

Vision not only detects and recognizes objects, but performs rich inferences about the underlying scene structure that causes the patterns of light we see. Inverting generative models, or "analysis-by-synthesis", presents a possible solution, but its mechanistic implementations have typically been too slow for online perception, and their mapping to neural circuits remains unclear. Here we present a neurally plausible efficient inverse graphics model and test it in the domain of face recognition. The model is based on a deep neural network that learns to invert a three-dimensional face graphics program in a single fast feedforward pass. It explains human behavior qualitatively and quantitatively, including the classic "hollow face" illusion, and it maps directly onto a specialized face-processing circuit in the primate brain. The model fits both behavioral and neural data better than state-of-the-art computer vision models, and suggests an interpretable reverse-engineering account of how the brain transforms images into percepts.

Modeling human intuitions about liquid flow with particle-based simulation.

Humans can easily describe, imagine, and, crucially, predict a wide variety of behaviors of liquids-splashing, squirting, gushing, sloshing, soaking, dripping, draining, trickling, pooling, and pouring-despite tremendous variability in their material and dynamical properties. Here we propose and test a computational model of how people perceive and predict these liquid dynamics, based on coarse approximate simulations of fluids as collections of interacting particles. Our model is analogous to a "game engine in the head", drawing on techniques for interactive simulations (as in video games) that optimize for efficiency and natural appearance rather than physical accuracy. In two behavioral experiments, we found that the model accurately captured people's predictions about how liquids flow among complex solid obstacles, and was significantly better than several alternatives based on simple heuristics and deep neural networks. Our model was also able to explain how people's predictions varied as a function of the liquids' properties (e.g., viscosity and stickiness). Together, the model and empirical results extend the recent proposal that human physical scene understanding for the dynamics of rigid, solid objects can be supported by approximate probabilistic simulation, to the more complex and unexplored domain of fluid dynamics.

An integrative computational architecture for object-driven cortex.

Computational architecture for object-driven cortex Objects in motion activate multiple cortical regions in every lobe of the human brain. Do these regions represent a collection of independent systems, or is there an overarching functional architecture spanning all of object-driven cortex? Inspired by recent work in artificial intelligence (AI), machine learning, and cognitive science, we consider the hypothesis that these regions can be understood as a coherent network implementing an integrative computational system that unifies the functions needed to perceive, predict, reason about, and plan with physical objects-as in the paradigmatic case of using or making tools. Our proposal draws on a modeling framework that combines multiple AI methods, including causal generative models, hybrid symbolic-continuous planning algorithms, and neural recognition networks, with object-centric, physics-based representations. We review evidence relating specific components of our proposal to the specific regions that comprise object-driven cortex, and lay out future research directions with the goal of building a complete functional and mechanistic account of this system.

Determining the Efficiency of Different Preoperative Difficult Intubation Tests on Patients Undergoing Caesarean Section.

BACKGROUND: Pregnancy-induced anatomical and physiological changes in the airway make airway management difficult in obstetric patients; thus, preoperative evaluation of the airway is important for obstetric patients. AIMS: To determine the effectiveness of the modified Mallampati test; the interincisor, sternomental and thyromental distances and the upper limb bite test. The second aim was to assess the effectiveness of the combination of the upper limb bite test with the other tests in obstetric patients. STUDY DESIGN: Cross-sectional study. METHODS: Pregnant women (n=250) scheduled for caesarean section were analysed. The patients' ages, heights and weights were collected. Preoperative airway evaluation was done by using a modified version of the Mallampati test. The interincisor, sternomental and thyromental distances were measured, and the upper limb bite test was performed. The laryngoscopy difficulty was evaluated by using Cormack-Lehane classification. RESULTS: No statistically significant differences were found between groups in age, height or weight (p>0.05). The modified Mallampati test and interincisor, sternomental and thyromental distances revealed a lower number of easy intubations than that determined by the Cormack-Lehane classification and a higher number of difficult intubations than the actual number of cases (p<0.05). The sensitivity and specificity of the modified Mallampati test, the upper limb bite test, the interincisor distance test and the sternomental and thyromental distance tests were found to be 73.08, 57.69, 84.62, 80.77 and 88.46 and 90.62, 99.11, 83.04, 84.37 and 87.05, respectively. When the combinations were examined, the sensitivity and specificity of the combination of the upper limb bite test with the modified Mallampati test were found to be 57.69 and 100, respectively. When the upper limb bite test was combined with the interincisor distance, the sensitivity and specificity were 46.15 and 100, respectively. We found a sensitivity and specificity of 93.75 and 95.30, respectively, for the combination of the upper limb bite test with the thyromental distance test. The sensitivity and specificity of the combination of the upper limb bite test with the modified Mallampati test and interincisor distance test were found to be 46.15 and 100, respectively. For combination of all the tests, the sensitivity and specificity was 42.31 and 100, respectively. CONCLUSION: When all combinations are evaluated in the decision of difficult intubation, the combination of the upper limb bite test and thyromental distance test is superior to the use of other methods alone to predict difficult intubation in pregnant women.

[The prognostic value of cerebral oxygen saturation measurement for assessing prognosis after cardiopulmonary resuscitation]. / O valor da medida da saturação cerebral de oxigênio para avaliar o prognóstico após ressuscitação cardiopulmonar.

BACKGROUND: Despite new improvements on cardiopulmonary resuscitation (CPR), brain damage is very often after resuscitation. OBJECTIVE: To assess the prognostic value of cerebral oxygen saturation measurement (rSO2) for assessing prognosis on patients after cardiopulmonary resuscitation. DESIGN: Retrospective analysis. MEASUREMENTS AND RESULTS: We analyzed 25 post-CPR patients (12 female and 13 male). All the patients were cooled to a target temperature of 33-34°C. The Glascow Coma Scale (GCS), Corneal Reflexes (CR), Pupillary Reflexes (PR), arterial Base Excess (BE) and rSO2 measurements were taken on admission. The rewarming GCS, CR, PR, BE and rSO2 measurements were made after the patient's temperature reached 36°C. RESULTS: In survivors, the baseline rSO2 value was 67.5 (46-70) and the percent difference between baseline and rewarming rSO2 value was 0.03 (0.014-0.435). In non-survivors, the baseline rSO2 value was 30 (25-65) and the percent difference between baseline and rewarming rSO2 value was 0.031 (-0.08 to -20). No statistical difference was detected on percent changes between baseline and rewarming values of rSO2. Statistically significant difference was detected between baseline and rewarming GCS groups (p=0.004). No statistical difference was detected between GCS, CR, PR, BE and rSO2 to determine the prognosis. CONCLUSION: Despite higher values of rSO2 on survivors than non-survivors, we found no statistically considerable difference between groups on baseline and the rewarming rSO2 values. Since the measurement is simple, and not affected by hypotension and hypothermia, the rSO2 may be a useful predictor for determining the prognosis after CPR.