Merlin: A Vision Language Foundation Model for 3D Computed Tomography.

Over 85 million computed tomography (CT) scans are performed annually in the US, of which approximately one quarter focus on the abdomen. Given the current shortage of both general and specialized radiologists, there is a large impetus to use artificial intelligence to alleviate the burden of interpreting these complex imaging studies while simultaneously using the images to extract novel physiological insights. Prior state-of-the-art approaches for automated medical image interpretation leverage vision language models (VLMs) that utilize both the image and the corresponding textual radiology reports. However, current medical VLMs are generally limited to 2D images and short reports. To overcome these shortcomings for abdominal CT interpretation, we introduce Merlin - a 3D VLM that leverages both structured electronic health records (EHR) and unstructured radiology reports for pretraining without requiring additional manual annotations. We train Merlin using a high-quality clinical dataset of paired CT scans (6+ million images from 15,331 CTs), EHR diagnosis codes (1.8+ million codes), and radiology reports (6+ million tokens) for training. We comprehensively evaluate Merlin on 6 task types and 752 individual tasks. The non-adapted (off-the-shelf) tasks include zero-shot findings classification (31 findings), phenotype classification (692 phenotypes), and zero-shot cross-modal retrieval (image to findings and image to impressions), while model adapted tasks include 5-year chronic disease prediction (6 diseases), radiology report generation, and 3D semantic segmentation (20 organs). We perform internal validation on a test set of 5,137 CTs, and external validation on 7,000 clinical CTs and on two public CT datasets (VerSe, TotalSegmentator). Beyond these clinically-relevant evaluations, we assess the efficacy of various network architectures and training strategies to depict that Merlin has favorable performance to existing task-specific baselines. We derive data scaling laws to empirically assess training data needs for requisite downstream task performance. Furthermore, unlike conventional VLMs that require hundreds of GPUs for training, we perform all training on a single GPU. This computationally efficient design can help democratize foundation model training, especially for health systems with compute constraints. We plan to release our trained models, code, and dataset, pending manual removal of all protected health information.

Assessing the Risk Stratification of Breast Cancer Polygenic Risk Scores in a Brazilian Cohort.

Polygenic risk scores (PRSs) for breast cancer have a clear clinical utility in risk prediction. PRS transferability across populations and ancestry groups is hampered by population-specific factors, ultimately leading to differences in variant effects, such as linkage disequilibrium and differences in variant frequency (allele frequency differences). Thus, locally sourced population-based phenotypic and genomic data sets are essential to assess the validity of PRSs derived from signals detected across populations. This study assesses the transferability of a breast cancer PRS composed of 313 risk variants (313-PRS) in a Brazilian trihybrid admixed ancestries (European, African, and Native American) whole-genome sequenced cohort, the Rare Genomes Project. 313-PRS was computed in the Rare Genomes Project (n = 853) using the UK Biobank (UKBB; n = 264,307) as reference. The Brazilian cohorts have a high European ancestry (EA) component, with allele frequency differences and to a lesser extent linkage disequilibrium patterns similar to those found in EA populations. The 313-PRS distribution was found to be inflated when compared with that of the UKBB, leading to potential overestimation of PRS-based risk if EA is taken as a standard. However, case controls lead to equivalent predictive power when compared with UKBB-EA samples with area under the receiver operating characteristic curve values of 0.66 to 0.62 compared with 0.63 for UKBB.

Automated abdominal CT contrast phase detection using an interpretable and open-source artificial intelligence algorithm.

OBJECTIVES: To develop and validate an open-source artificial intelligence (AI) algorithm to accurately detect contrast phases in abdominal CT scans. MATERIALS AND METHODS: Retrospective study aimed to develop an AI algorithm trained on 739 abdominal CT exams from 2016 to 2021, from 200 unique patients, covering 1545 axial series. We performed segmentation of five key anatomic structures-aorta, portal vein, inferior vena cava, renal parenchyma, and renal pelvis-using TotalSegmentator, a deep learning-based tool for multi-organ segmentation, and a rule-based approach to extract the renal pelvis. Radiomics features were extracted from the anatomical structures for use in a gradient-boosting classifier to identify four contrast phases: non-contrast, arterial, venous, and delayed. Internal and external validation was performed using the F1 score and other classification metrics, on the external dataset "VinDr-Multiphase CT". RESULTS: The training dataset consisted of 172 patients (mean age, 70 years ± 8, 22% women), and the internal test set included 28 patients (mean age, 68 years ± 8, 14% women). In internal validation, the classifier achieved an accuracy of 92.3%, with an average F1 score of 90.7%. During external validation, the algorithm maintained an accuracy of 90.1%, with an average F1 score of 82.6%. Shapley feature attribution analysis indicated that renal and vascular radiodensity values were the most important for phase classification. CONCLUSION: An open-source and interpretable AI algorithm accurately detects contrast phases in abdominal CT scans, with high accuracy and F1 scores in internal and external validation, confirming its generalization capability. CLINICAL RELEVANCE STATEMENT: Contrast phase detection in abdominal CT scans is a critical step for downstream AI applications, deploying algorithms in the clinical setting, and for quantifying imaging biomarkers, ultimately allowing for better diagnostics and increased access to diagnostic imaging. KEY POINTS: Digital Imaging and Communications in Medicine labels are inaccurate for determining the abdominal CT scan phase. AI provides great help in accurately discriminating the contrast phase. Accurate contrast phase determination aids downstream AI applications and biomarker quantification.

Medical Expectations of Physicians on AI Solutions in Daily Practice: Cross-Sectional Survey Study.

Background: The use of artificial intelligence (AI) in medicine has been a trending subject in the past few years. Although not frequently used in daily practice yet, it brings along many expectations, doubts, and fears for physicians. Surveys can be used to help understand this situation. Objective: This study aimed to explore the degree of knowledge, expectations, and fears on possible AI use by physicians in daily practice, according to sex and time since graduation. Methods: An electronic survey was sent to physicians of a large hospital in Brazil, from August to September 2022. Results: A total of 164 physicians responded to our survey. Overall, 54.3% (89/164) of physicians considered themselves to have an intermediate knowledge of AI, and 78.5% (128/163) believed that AI should be regulated by a governmental agency. If AI solutions were reliable, fast, and available, 77.9% (127/163) intended to frequently or always use AI for diagnosis (143/164, 87.2%), management (140/164, 85.4%), or exams interpretation (150/164, 91.5%), but their approvals for AI when used by other health professionals (85/163, 52.1%) or directly by patients (82/162, 50.6%) were not as high. The main benefit would be increasing the speed for diagnosis and management (106/163, 61.3%), and the worst issue would be to over rely on AI and lose medical skills (118/163, 72.4%). Physicians believed that AI would be useful (106/163, 65%), facilitate their work (140/153, 91.5%), not alter the number of appointments (80/162, 49.4%), not interfere in their financial gain (94/162, 58%), and not replace their jobs but be an additional source of information (104/162, 64.2%). In case of disagreement between AI and physicians, most (108/159, 67.9%) answered that a third opinion should be requested. Physicians with ≤10 years since graduation would adopt AI solutions more frequently than those with >20 years since graduation (P=.04), and female physicians were more receptive to other hospital staff using AI than male physicians (P=.008). Conclusions: Physicians were shown to have good expectations regarding the use of AI in medicine when they apply it themselves, but not when used by others. They also intend to use it, as long as it was approved by a regulatory agency. Although there was hope for a beneficial impact of AI on health care, it also brings specific concerns.

Left-sided epileptiform activity influences language lateralization in right mesial temporal sclerosis.

OBJECTIVE: To investigate the association between left epileptiform activity and language laterality indices (LI) in patients with right mesial temporal sclerosis (MTS). METHODS: Twenty-two patients with right MTS and 22 healthy subjects underwent fMRI scanning while performing a language task. LI was calculated in multiple regions of interest (ROI). Data on the presence of left epileptiform abnormalities were obtained during prolonged video-EEG monitoring. RESULTS: After correction for multiple comparisons, LI was reduced in the middle temporal gyrus in the left interictal epileptiform discharges (IED+) group, compared with the left IED- group (p < 0.05). SIGNIFICANCE: Using a responsive reading naming fMRI paradigm, right MTS patients who presented left temporal interictal epileptiform abnormalities on video-EEG showed decreased LI in the middle temporal gyrus, indicating decreased left middle temporal gyrus activation, increased right middle temporal gyrus activation or a combination of both, demonstrative of language network reorganization, specially in the MTG, in this patient population. PLAIN LANGUAGE SUMMARY: This research studied 22 patients with right mesial temporal sclerosis (a specific type of epilepsy) comparing them to 22 healthy individuals. Participants were asked to perform a language task while undergoing a special brain imaging technique (fMRI). The findings showed that patients with epilepsy displayed a change in the area of the brain typically responsible for language processing. This suggests that their brains may have adapted due to their condition, altering the way language is processed.

Perilesional and contralesional brain activations related to associative encoding of unfamiliar face-names pairs in adults with left chronic stroke with or without ischemic infarct on left inferior frontal gyrus.

The study of an Ischemic stroke infarction allows verifying how the lesion produces alterations in the neuronal networks resulting in cognitive deficits. It also allows the verification of adaptive and maladaptive cerebral reorganization related to the injury. In our previous fMRI study, we found that patients without ischemic vascular lesions in left inferior frontal gyrus showed an efficient compensation mechanism during the associative encoding of face name pairs, by the increased activation of ventrolateral and dorsolateral areas of contralesional hemisphere associated with better memory performance. While patients with ischemic vascular lesions on left inferior frontal gyrus (IFG) demonstrated worse memory performance and no signs of compensation mechanism. The present study explores more of these findings by analyzing perilesional and contralesional activations related to unfamiliar face name associative encoding in adults with chronic ischemic stroke, with or without left IFG lesion, compared to healthy controls. The main results showed that stroke survivors without lesions in IFG demonstrated increased activation in perilesional and contralesional prefrontal regions associated with better associative memory recognition, which are indicative of adaptive compensatory mechanisms. However, they also showed a negative correlation between the activation of right anterior prefrontal and inferior parietal regions and the associative memory performance, which may indicate the presence of maladaptive interhemispheric disinhibition. On the other hand, stroke survivors with IFG lesions demonstrated negative correlations in activations of the ipsilesional inferior parietal cortex and positive correlations in activations of the left middle frontal gyrus and left precentral cortex, which demonstrate the simultaneous occurrence of adaptive and maladaptive brain reorganization mechanisms in this group. However, the increase in perilesional prefrontal regions, associated with bilateral activation of the hippocampus and amygdala, was not enough to compensate for the inefficiency of associative memory performance. Finally, the differences in activation observed in stroke survivors reflect their clinical heterogeneity and demonstrate that adaptive or maladaptive compensatory mechanisms can coexist in the same group of patients. Furthermore, they reinforce the importance of the left IFG in the associative encoding of unfamiliar face name pairs and may suggest a deficit in associative memory related to injury in this region.

The Impact of Post-Stroke Depressive Symptoms on Cognitive Performance in Women and in Men: A 4 Month Prospective Study.

Background: Depressive symptoms have been associated with cognitive impairment after stroke, and women may be specifically affected. Objective: The aim of this study was to investigate gender-specific characteristics in the relationship between changes in depression severity and changes in cognitive performance after stroke. Methods: We prospectively evaluated 73 patients without a previous history of depression in the first and fourth months after a first ischemic stroke. The severity of depressive symptoms was assessed using the 31-item version of the Hamilton Rating Scale for Depression, and executive function, attention, working memory, and verbal fluency were assessed using a neuropsychological battery. Results: We included 46 (63.0%) men and 27 (36.9%) women, with mean ages of 55.2 (SD ± 15.1) and 46.8 (SD ± 14.7) years, respectively. We found significant improvement in the digit span forward and Stroop dots from month 1 to month 4 post stroke for both men and women. Women, but not men, presented a correlation between changes in phonemic verbal fluency and changes in the 31-item version of the Hamilton Rating Scale for Depression scores. Improvement in depression was correlated with improvement in verbal fluency, and worsening in depression was correlated with worsening in verbal fluency. Conclusions: Our results suggest that women might be more vulnerable to the relationship between depressive symptoms and cognitive performance, and improvement of depression may be necessary for women's improvement in phonemic verbal fluency from the first to the fourth month after a stroke. We did not adjust the results for multiple comparisons. Thus, our findings might be considered preliminary, and confirmatory studies, also focusing on specific characteristics of women that could explain these differences, are warranted.

Human motor sequence learning drives transient changes in network topology and hippocampal connectivity early during memory consolidation.

In the last decade, the exclusive role of the hippocampus in human declarative learning has been challenged. Recently, we have shown that gains in performance observed in motor sequence learning (MSL) during the quiet rest periods interleaved with practice are associated with increased hippocampal activity, suggesting a role of this structure in motor memory reactivation. Yet, skill also develops offline as memory stabilizes after training and overnight. To examine whether the hippocampus contributes to motor sequence memory consolidation, here we used a network neuroscience strategy to track its functional connectivity offline 30 min and 24 h post learning using resting-state functional magnetic resonance imaging. Using a graph-analytical approach we found that MSL transiently increased network modularity, reflected in an increment in local information processing at 30 min that returned to baseline at 24 h. Within the same time window, MSL decreased the connectivity of a hippocampal-sensorimotor network, and increased the connectivity of a striatal-premotor network in an antagonistic manner. Finally, a supervised classification identified a low-dimensional pattern of hippocampal connectivity that discriminated between control and MSL data with high accuracy. The fact that changes in hippocampal connectivity were detected shortly after training supports a relevant role of the hippocampus in early stages of motor memory consolidation.

Long term effects of red wine consumption in brain: an MRI, fMRI and neuropsychological evaluation study.

Red wine (RW) consumption has been proposed to have a potential health benefit. However, the effect of RW consumption on the brain is not entirely known, mainly when associated with aging. Regular red wine consumers (n = 30) and abstainers (ABST; n = 27) without cognitive impairment were evaluated for brain structural characteristics (Fazekas score and voxel-based morphometry) and for functional adaptations assessed by fMRI (using the Word Tasks Color Stroop (WCST) and Two-Back (TBT)), as well as by neuropsychological tests in different domains. There were no significant differences regarding brain morphological features. RW consumers showed greater activation in the thalamus during WCST and in paracingulate/anterior cingulate cortices, left superior frontal gyrus and frontal pole during TBT. ABST required higher activation of different cortical areas in the left parietal lobe during WCST. Age and intelligence quotient influenced those activations. In Stroop and trail-making neuropsychological tests, RW consumers performed slightly better than ABST. This study should be viewed as hypothesis-generating rather than conclusive.HighlightsWhite matter hyperintensities and gray matter volume did not differ between the RW and ABST groups.RW consumers could depend more on right thalamus during WSCT due to its role in visual integration.ABST could depend more on left parietal lobe during WSCT due to its role in sensory and phonological encoding.RW consumers with inferior cognitive abilities could depend more on letter recognition to solve a TBT correctly.Younger abstainers could depend more on different areas involved in integrating cognitive processes and attention regulation to solve a TBT correctly.

Nationwide Cross-sectional Analysis of Endoscopic Thoracic Sympathectomy to Treat Hyperhidrosis Over 12 years in Brazil: Epidemiology, Costs, and Mortality.

OBJECTIVE: To analyze the number of endoscopic thoracic sympathectomies performed to treat hyperhidrosis in the Universal Public Health System of Brazil, the government reimbursements, and the in-hospital mortality rates. BACKGROUND: Even though endoscopic thoracic sympathectomy has been widely performed for the definitive treatment of hyperhidrosis, no series reported mortality and there are no population-based studies evaluating its costs or its mortality rate. METHODS: Data referring to endoscopic thoracic sympathectomy to treat hyperhidrosis between 2008 and 2019 were extracted from the database of the Brazilian Public Health System, which insures more than 160 million inhabitants. RESULTS: Thirteen thousand two hundred one endoscopic thoracic sympathectomies to treat hyperhidrosis were performed from 2008 to 2019, with a rate of 68.44 procedures per 10 million inhabitants per year. There were 6 in-hospital deaths during the whole period, representing a mortality rate of 0.045%. The total expended throughout the years was U$ 6,767,825.14, with an average of U$ 512.68 per patient. CONCLUSIONS: We observed a rate of 68.44 thoracoscopic sympathectomies for hyperhidrosis' treatment per 10 million inhabitants per year. The inhospital mortality rate was very low, 0.045%, though not nil. To our knowledge, no published series is larger than ours and we are the first authors to formally report deaths after endoscopic thoracic sympathectomies to treat hyperhidrosis. Moreover, there is no other population-based study addressing costs and mortality rates of every endoscopic thoracic sympathectomy for the treatment of any site of hyperhidrosis in a given period.

Freezing of gait, gait initiation, and gait automaticity share a similar neural substrate in Parkinson's disease.

Individuals with Parkinson's disease (PD) and freezing of gait (FOG) have difficulty initiating and maintaining a healthy gait pattern; however, the relationship among FOG severity, gait initiation, and gait automaticity, in addition to the neural substrate of this relationship has not been investigated. This study investigated the association among FOG severity during turning (FOG-ratio), gait initiation (anticipatory postural adjustment [APA]), and gait automaticity (dual-task cost [DTC]), and the neural substrates of these associations. Thirty-four individuals with FOG of PD were assessed in the ON-medication state. FOG-ratio during a turning test, gait automaticity using DTC on stride length and gait speed, and APA during an event-related functional magnetic resonance imaging protocol to assess brain activity from the regions of interest (e.g., dorsolateral prefrontal cortex [DLPFC] and mesencephalic locomotor region [MLR]) were assessed in separated days. Results showed that FOG-ratio, APA amplitude, and DTC on stride length are negatively associated among them (P < 0.05). APA amplitude and DTC on stride length explained 59% of the FOG-ratio variance (P < 0.05). Although the activity of the right DLPFC and right MLR explained 55% of the FOG-ratio variance (P < 0.05) and 30% of the DTC on stride length variance (P ≤ 0.05), only the activity of the right MLR explained 23% of the APA amplitude (P < 0.05). FOG severity during turning, APA amplitude, and stride length automaticity are associated among them and share a similar locomotor substrate, as the MLR activity was a common brain region in explaining the variance of these variables.

Epidemiological analysis of lower limb revascularization for peripheral arterial disease over 12 years on the public healthcare system in Brazil.

Background: Worldwide, peripheral arterial disease (PAD) is a disorder with high morbidity, affecting more than 200 million people. Objectives: Our objective was to analyze surgical treatment for PAD provided on the Brazilian Public Healthcare System over 12 years using publicly available data. Methods: The study was conducted with analysis of data available on the Brazilian Health Ministry's database platform, assessing distributions of procedures and techniques over the years and their associated mortality and costs. Results: A total of 129,424 procedures were analyzed (performed either for claudication or critical ischemia, proportion unknown). The vast majority of procedures were endovascular (65.49%) and this disproportion exhibited a rising trend (p<0.001). There were 3,306 in-hospital deaths (mortality of 2.55%), with lower mortality in the endovascular group (1.2% vs. 5.0%, p=0.008). The overall governmental expenditure on these procedures was U$ 238,010,096.51, and endovascular procedures were on average significantly more expensive than open surgery (U$ 1,932.27 vs. U$ 1,517.32; p=0.016). Conclusions: Lower limb revascularizations were performed on the Brazilian Public Healthcare System with gradually increasing frequency from 2008 to 2019. Endovascular procedures were vastly more common and were associated with lower in-hospital mortality rates, but higher procedure costs.

Contexto: A doença arterial periférica (DAP) é uma doença com alta morbidade global, afetando mais de 200 milhões de pessoas. Objetivos: Neste estudo, analisamos o tratamento cirúrgico para DAP no sistema público de saúde do Brasil no período de 12 anos, com base em dados publicamente disponíveis. Métodos: O estudo foi conduzido a partir da análise de dados disponíveis na plataforma do Departamento de Informática do Sistema Único de Saúde (DATASUS), do Ministério da Saúde, avaliando a distribuição da técnica cirúrgica utilizada, a mortalidade e o custo ao longo dos anos. Resultados: Um total de 129.424 procedimentos foram analisados (para claudicantes e isquemia crítica, em proporção desconhecida). A maiora dos procedimentos foi via endovascular (65,49%), com tendência de aumento nessa desproporção (p < 0,001). Houve 3.306 mortes intra-hospitalares (mortalidade de 2,55%) com menor mortalidade no grupo endovascular (1,2% vs. 5,0%; p = 0,008). O investimento governamental total para esses procedimentos foi de US$ 238.010.096,51, e os procedimentos endovasculares foram significativamente mais caros que a cirurgia aberta convencional (US$ 1.932,27 vs. US$ 1.517,32; p = 0,016). Conclusões: No sistema público de saúde brasileiro, as revascularizações de membros inferiores ocorreram com frequência crescente entre 2008 e 2019. Os procedimentos endovasculares foram mais comuns e relacionados a menor mortalidade intra-hospitalar, mas a maiores custos.

BRAX, Brazilian labeled chest x-ray dataset.

Chest radiographs allow for the meticulous examination of a patient's chest but demands specialized training for proper interpretation. Automated analysis of medical imaging has become increasingly accessible with the advent of machine learning (ML) algorithms. Large labeled datasets are key elements for training and validation of these ML solutions. In this paper we describe the Brazilian labeled chest x-ray dataset, BRAX: an automatically labeled dataset designed to assist researchers in the validation of ML models. The dataset contains 24,959 chest radiography studies from patients presenting to a large general Brazilian hospital. A total of 40,967 images are available in the BRAX dataset. All images have been verified by trained radiologists and de-identified to protect patient privacy. Fourteen labels were derived from free-text radiology reports written in Brazilian Portuguese using Natural Language Processing.

Surgical repair of abdominal aortic aneurysms on the public health system in the largest city in Brazil: a descriptive analysis of in-hospital data on 2693 procedures over 10 years.

Background: From 1990 to 2015, mortality from aortic aneurysms increased 16.8% in Brazil. São Paulo is the largest city in Brazil and about 5 million people depend on the public health system there. Objectives: To conduct an epidemiological analysis of abdominal aortic aneurysm surgeries in the city of São Paulo. Methods: Infra-renal aortic aneurysm procedures performed over a decade (from 2008 to 2017) were studied using publicly-available platforms from the Unified Health System and DATASUS. Results: 2693 procedures were analyzed; 66.73% were endovascular; 78.7% of patients were male; 70.7% were aged 65 years or more; 64.02% were elective hospital admissions. There were 288 in-hospital deaths (mortality: 10.69%). In-hospital mortality was lower for endovascular surgery than for open surgery; both for elective (4.13% versus 14.42%) and urgent (9.73% versus 27.94%) (p = 0.019) admissions. The highest volume hospital (n = 635) had the lowest in-hospital mortality (3.31%). USD 24,835,604.84 was paid; an average of $ 2,318.63 for elective open, $ 3,420.10 for emergency open, $ 12,157.35 for elective endovascular and $ 12,969.12 for urgent endovascular procedures. Endovascular procedure costs were statistically higher than the values paid for open surgeries (p <0.001). Conclusions: Endovascular surgeries were performed twice as often as open surgeries; they had shorter hospital stays and lower mortality.

Contexto: No Brasil, a mortalidade por aneurisma de aorta aumentou 16,8% de 1990 a 2015. São Paulo é a maior cidade do Brasil, e cerca de 5 milhões de pessoas dependem do sistema público de saúde. Objetivos: Análise epidemiológica das cirurgias do aneurisma de aorta abdominal na cidade de São Paulo. Métodos: As cirurgias para correção do aneurisma de aorta infrarrenal realizadas no período de uma década (de 2008 a 2017) foram estudadas utilizando-se plataformas publicamente disponíveis do Sistema Único de Saúde e do Departamento de Informática do Sistema Único de Saúde. Resultados: Foram analisados ​​2.693 procedimentos, entre os quais 66,73% eram endovasculares. Entre os pacientes, houve predominância do sexo masculino (78,7%) e daqueles com 65 anos ou mais (70,7%). Um total de 64,02% eram admissões hospitalares eletivas. Ocorreram 288 óbitos hospitalares (mortalidade: 10,69%). A mortalidade durante a internação foi menor para cirurgia endovascular do que para cirurgia aberta tanto no contexto eletivo (4,13% versus 14,42%) quanto urgente (9,73% versus 27,94%) (p = 0,019). O maior volume (n = 635) apresentou menor mortalidade intra-hospitalar (3,31%). Foi pago um total de $24.835.604,84, sendo uma média de $2.318,63 para cirurgia abertura eletiva, $3.420,10 para cirurgia abertura de emergência, $12.157,35 para cirurgia endovascular eletiva e $12.969,12 para cirurgia endovascular na urgência. Os custos dos procedimentos endovasculares foram estatisticamente superiores aos valores pagos para as cirurgias abertas (p < 0,001). Conclusões: Foram realizadas duas vezes mais cirurgias endovasculares do que abertas, as quais apresentaram menor tempo de internação e menor mortalidade.

Prediction of metabolic syndrome: A machine learning approach to help primary prevention.

AIMS: To describe the performance of machine learning (ML) applied to predict future metabolic syndrome (MS), and to estimate lifestyle changes effects in MS predictions. METHODS: We analyzed data from 17,182 adults attending a checkup program sequentially (37,999 visit pairs) over 17 years. Variables on sociodemographic attributes, clinical, laboratory, and lifestyle characteristics were used to develop ML models to predict MS [logistic regression, linear discriminant analysis, k-nearest neighbors, decision trees, Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting]. We have tested the effects of changes in lifestyle in MS prediction at individual levels. RESULTS: All models showed adequate calibration and good discrimination, but the LGBM showed better performance (Sensitivity = 87.8 %, Specificity = 70.2 %, AUC-ROC = 0.86). Causal inference analysis showed that increasing physical activity level and reducing BMI by at least 2 % had an effect of reducing the predicted probability of MS by 3.8 % (95 % CI = -4.8 %; -2.7 %). CONCLUSION: ML models based on data from a checkup program showed good performance to predict MS and allowed testing for effects of lifestyle changes in this prediction. External validation is recommended to verify models' ability to identify at-risk individuals, and potentially increase their engagement in preventive measures.

Epidemiological Analysis of 12 Years of Open Thoracoabdominal Aortic Aneurysm Repair in the Brazilian Public Health System.

INTRODUCTION: Although endovascular correction is a promising perspective, the gold-standard treatment for thoracoabdominal aortic aneurisms and type-B dissections with visceral involvement remains open surgery, particularly due to its well-established long-term durability. This study aims to describe and evaluate public data from patients treated for thoracoabdominal aortic aneurism in the Brazilian public health system in a 12-year interval. METHODS: Data from procedures performed between 2008 and 2019 were extracted from the national public database (Departamento de Informática do Sistema Único de Saúde, or DATASUS) using web scraping techniques. Procedures were evaluated regarding the yearly frequency of elective or urgency surgeries, in-hospital mortality, and governmental costs. All tests were done with a level of significance P<0.05. RESULTS: A total of 812 procedures were analyzed. Of all surgeries, 67.98% were elective cases. There were 328 in-hospital deaths (mortality of 40.39%). In-hospital mortality was lower in elective procedures (26.92%) than in urgency procedures (46.74%) (P=0.008). Total governmental expenditure was $3.127.051,56 - an average of $3.774,22 for elective surgery and $3.791,93 for emergency surgery (P=0.999). CONCLUSION: The proportion of urgency procedures is higher than that recommended by international literature. Mortality was higher for urgent admissions, although governmental costs were equal for elective and urgent procedures; specialized referral centers should be considered by health policy makers.

Temporal trends in vena cava filter implantation in public health system inpatients: an 11-year analysis of the largest city in Brazil.

Background: Vena cava filter implantation is considered a simple procedure, which can lead to overuse and over-indication. It is nevertheless associated with short and long-term complications. Objectives: The goals of this study were to evaluate rates of vena cava filter implantation conducted by Brazil's Unified Public Health System, analyzing in-hospital mortality and migration of patients from other cities seeking medical attention in São Paulo. Methods: This study analyzed all vena cava filter procedures conducted from 2008 to 2018 in the city of São Paulo and registered on the public database using a big data system to conduct web scraping of publicly available databases. Results: A total of 1324 vena cava filter implantations were analyzed. 60.5% of the patients were female; 61.7% were under 65 years old; 34.07% had registered addresses in other cities or states; and there was a 7.4% in-hospital mortality rate. Conclusions: We observed an increase in the rates of use of vena cava filters up to 2010 and a decrease in rates from that year onwards, which coincides with the year that the Food and Drug Administration published a recommendation to better evaluate vena cava filter indications.

Contexto: O implante de filtro de veia cava é considerado um procedimento de baixa complexidade, o que pode resultar em indicação excessiva. No entanto, não é isento de complicações a curto e longo prazo. Objetivos: Avaliar as taxas de implantes de filtro de veia cava realizados pelo Sistema Único de Saúde e a origem geográfica e mortalidade intra-hospitalar dos pacientes. Métodos: Foi conduzida uma análise em um banco de dados públicos referente às taxas de implantes de filtro de veia cava realizados de 2008 a 2018 na cidade de São Paulo, utilizando o sistema de big data. Resultados: Foram analisados 1.324 implantes de filtro de veia cava financiados pelo Sistema Único de Saúde. Identificou-se tendência de aumento da taxa de implantação até 2010 e de redução dos números após esse período. Do total de pacientes, 60,5% eram do sexo feminino; 61,75% tinham menos de 65 anos; e 34,07% possuíam endereço oficial em outra cidade ou estado. A taxa de mortalidade intra-hospitalar foi de 7,4%. Conclusões: Observamos aumento das taxas de implante de filtro de veia cava até 2010 e redução das taxas após esse período, o que coincide com o ano em que a organização norte-americana Food and Drug Administration publicou uma recomendação para melhor avaliar as indicações de filtros.

Effects of Repetitive Peripheral Sensory Stimulation in the Subacute and Chronic Phases After Stroke: Study Protocol for a Pilot Randomized Trial.

BACKGROUND: Repetitive peripheral nerve sensory stimulation (RPSS) is a potential add-on intervention to motor training for rehabilitation of upper limb paresis after stroke. Benefits of RPSS were reported in subjects in the chronic phase after stroke, but there is limited information about the effects of this intervention within the 1st weeks or months. The primary goal of this study is to compare, in a head-to-head proof-of-principle study, the impact of a single session of suprasensory vs. subsensory RPSS on the upper limb motor performance and learning in subjects at different phases after stroke subacute and chronic phases and mild upper limb motor impairments after stroke. In addition, we examine the effects of RPSS on brain perfusion, functional imaging activation, and γ-aminobutyric acid (GABA) levels. Subjects with mild upper limb motor impairments will be tested with MRI and clinical assessment either at an early (7 days to 3 months post-stroke) or at a chronic (>6 months) stage after stroke. METHODS: In this multicenter, randomized, parallel-group, proof-of-principle clinical trial with blinded assessment of outcomes, we compare the effects of one session of suprasensory or subsensory RPSS in patients with ischemic or hemorrhagic stroke and upper limb paresis. Clinical assessment and MRI will be performed only once in each subject (either at an early or at a chronic stage). The primary outcome is the change in performance in the Jebsen-Taylor test. Secondary outcomes: hand strength, cerebral blood flow assessed with arterial spin labeling, changes in the blood oxygenation level-dependent (BOLD) effect in ipsilesional and contralesional primary motor cortex (M1) on the left and the right hemispheres assessed with functional MRI (fMRI) during a finger-tapping task performed with the paretic hand, and changes in GABA levels in ipsilesional and contralesional M1 evaluated with spectroscopy. The changes in outcomes will be compared in four groups: suprasensory, early; subsensory, early; suprasensory, chronic; and subsensory, chronic. DISCUSSION: The results of this study are relevant to inform future clinical trials to tailor RPSS to patients more likely to benefit from this intervention. TRIAL REGISTRATION: NCT03956407.

Epidemiology of Abdominal Aortic Aneurysm Repair in Brazil from 2008 to 2019 and Comprehensive Review of Nationwide Statistics Across the World.

INTRODUCTION: Studying epidemiology of abdominal aortic aneurysms repairs is essential to prevent related deaths. Although outcomes are influenced by socioeconomic factors, there are no nationwide studies on these statistics in low-and-middle income countries. Therefore, we designed this study to evaluate abdominal aortic aneurysms repair rates, trends, costs, and in-hospital mortality in the Brazilian Public Health System, which exclusively insures over 160 million Brazilians. MATERIAL AND METHODS: Retrospective cross-sectional population-based analysis of publicly available data referring to all abdominal aortic aneurysm repairs performed between 2008 and 2019 in Brazilian public hospitals. RESULTS: We observed a total of 13,506 abdominal aortic aneurysm repairs, of which 32% were emergency endovascular repairs, 20% emergency open repairs, 32% elective endovascular repairs and 16% elective open repairs. There has been a downward trend in total abdominal aortic aneurysms repairs and an increasing predominance of endovascular repair. Elective and endovascular repairs were significantly associated with lower in-hospital mortality. For ruptured abdominal aortic aneurysms, we observed mortality rates of 13.8% after endovascular repair and 52.1% after open repair. For intact abdominal aortic aneurysms, we observed mortalities of 3.8% after endovascular repair and 18.6% after open repair. Procedure and mortality rates varied significantly among the Brazilian regions. CONCLUSIONS: We observed a low and decreasing rate of abdominal aortic aneurysm repair. Most repairs were emergency and endovascular and there was an increasing predominance of endovascular repair. Endovascular and elective repairs were associated with lower mortality.

Corrigendum: Behavioral and Neural Correlates of Cognitive Training and Transfer Effects in Stroke Patients.

[This corrects the article DOI: 10.3389/fneur.2020.01048.].