Background: Machine learning (ML) has made a significant impact in medicine and cancer research; however, its impact in these areas has been undeniably slower and more limited than in other application domains. A major reason for this has been the lack of availability of patient data to the broader ML research community, in large part due to patient privacy protection concerns. High-quality, realistic, synthetic datasets can be leveraged to accelerate methodological developments in medicine. By and large, medical data is high dimensional and often categorical. These characteristics pose multiple modeling challenges. Methods: In this paper, we evaluate three classes of synthetic data generation approaches; probabilistic models, classification-based imputation models, and generative adversarial neural networks. Metrics for evaluating the quality of the generated synthetic datasets are presented and discussed. Results: While the results and discussions are broadly applicable to medical data, for demonstration purposes we generate synthetic datasets for cancer based on the publicly available cancer registry data from the Surveillance Epidemiology and End Results (SEER) program. Specifically, our cohort consists of breast, respiratory, and non-solid cancer cases diagnosed between 2010 and 2015, which includes over 360,000 individual cases. Conclusions: We discuss the trade-offs of the different methods and metrics, providing guidance on considerations for the generation and usage of medical synthetic data.

A flexible 3D microelectrode array (3DMEA) device was developed that non-invasively interrogates electrophysiology of 3D in vitro neuronal cultures.

The brain’s extracellular matrix (ECM) is a macromolecular network composed of glycosaminoglycans, proteoglycans, glycoproteins, and fibrous proteins. In vitro studies often use purified ECM proteins for cell culture coatings, however these may not represent the molecular complexity and heterogeneity of the brain’s ECM. To address this, we compared neural network activity (over 30 days in vitro) from primary neurons co-cultured with glia grown on ECM coatings from decellularized brain tissue (bECM) or MaxGel, a non-tissue-specific ECM. Cells were grown on a multi-electrode array (MEA) to enable noninvasive long-term interrogation of neuronal networks. In general, the presence of ECM accelerated the formation of networks without affecting the inherent network properties. However, specific features of network activity were dependent on the type of ECM: bECM enhanced network activity over a greater region of the MEA whereas MaxGel increased network burst rate associated with robust synaptophysin expression. These differences in network activity were not attributable to cellular composition, glial proliferation, or astrocyte phenotypes, which remained constant across experimental conditions. Collectively, the addition of ECM to neuronal cultures represents a reliable method to accelerate the development of mature neuronal networks, providing a means to enhance throughput for routine evaluation of neurotoxins and novel therapeutics.

In vitro brain-on-a-chip platforms hold promise in many areas including: drug discovery, evaluating effects of toxicants and pathogens, and disease modelling. A more accurate recapitulation of the intricate organization of the brain in vivo may require a complex in vitro system including organization of multiple neuronal cell types in an anatomically-relevant manner. Most approaches for compartmentalizing or segregating multiple cell types on microfabricated substrates use either permanent physical surface features or chemical surface functionalization. This study describes a removable insert that successfully deposits neurons from different brain areas onto discrete regions of a microelectrode array (MEA) surface, achieving a separation distance of 100 μm. The regional seeding area on the substrate is significantly smaller than current platforms using comparable placement methods. The non-permanent barrier between cell populations allows the cells to remain localized and attach to the substrate while the insert is in place and interact with neighboring regions after removal. The insert was used to simultaneously seed primary rodent hippocampal and cortical neurons onto MEAs. These cells retained their morphology, viability, and function after seeding through the cell insert through 28 days in vitro (DIV). Co-cultures of the two neuron types developed processes and formed integrated networks between the different MEA regions. Electrophysiological data demonstrated characteristic bursting features and waveform shapes that were consistent for each neuron type in both mono- and co-culture. Additionally, hippocampal cells co-cultured with cortical neurons showed an increase in within-burst firing rate (p = 0.013) and percent spikes in bursts (p = 0.002), changes that imply communication exists between the two cell types in co-culture. The cell seeding insert described in this work is a simple but effective method of separating distinct neuronal populations on microfabricated devices, and offers a unique approach to developing the types of complex in vitro cellular environments required for anatomically-relevant brain-on-a-chip devices.

Quantitatively benchmarking similarities and differences between the in vivo central nervous system and in vitro neuronal cultures can qualify discrepancies in functional responses and establish the utility of in vitro platforms. In this work, extracellular electrophysiology responses of cortical neurons in awake, freely-moving animals were compared to in vitro cultures of dissociated cortical neurons. After exposure to two well-characterized drugs, atropine and ketamine, a number of key points were observed: (1) significant differences in spontaneous firing activity for in vivo and in vitro systems, (2) similar response trends in single-unit spiking activity after exposure to atropine, and (3) greater sensitivity to the effects of ketamine in vitro. While in vitro cultures of dissociated cortical neurons may be appropriate for many types of pharmacological studies, we demonstrate that for some drugs, such as ketamine, this system may not fully capture the responses observed in vivo. Understanding the functionality associated with neuronal cultures will enhance the relevance of electrophysiology data sets and more accurately frame their conclusions. Comparing in vivo and in vitro rodent systems will provide the critical framework necessary for developing and interpreting in vitro systems using human cells that strive to more closely recapitulate human in vivo function and response.

Brain-on-a-chip systems are designed to simulate brain activity using traditional in vitro cell culture on an engineered platform. it is a noninvasive tool to screen new drugs, evaluate toxicants, and elucidate disease mechanisms. However, successful recapitulation of brain function on these systems is dependent on the complexity of the cell culture. in this study, we increased cellular complexity of traditional (simple) neuronal cultures by co-culturing with astrocytes and oligodendrocyte precursor cells (complex culture). We evaluated and compared neuronal activity (e.g., network formation and maturation), cellular composition in long-term culture, and the transcriptome of the two cultures. compared to simple cultures, neurons from complex co-cultures exhibited earlier synapse and network development and maturation, which was supported by localized synaptophysin expression, up-regulation of genes involved in mature neuronal processes, and synchronized neural network activity. Also, mature oligodendrocytes and reactive astrocytes were only detected in complex cultures upon transcriptomic analysis of age-matched cultures. functionally, the GABA antagonist bicuculline had a greater influence on bursting activity in complex versus simple cultures. Collectively, the cellular complexity of brain-on-a-chip systems intrinsically develops cell type-specific phenotypes relevant to the brain while accelerating the maturation of neuronal networks, important features underdeveloped in traditional cultures. In vitro brain-on-a-chip platforms have emerged as useful tools to model brain activity to aid in evaluating neuronal outcomes for new drugs and toxicants, in addition to elucidating disease mechanisms 1-3. These in vitro approaches often utilize multi-electrode arrays (MEA), which allow for non-invasive interrogation of in vitro neuronal networks formed de novo from dissociated rodent or human neurons or from networks established in rodent brain tissue slices 4-7. The use of dissociated neurons offers an amenable approach for establishing and evaluating human-relevant responses using human primary or stem-cell derived neurons and glial cell types 7-11 , since human brain slices are not often available. Brain-on-a-chip efforts incorporating either rodent or human cell types have been used for toxicology screening 12-14 , developing integrated systems (i.e. neurovascular units comprised of a blood-brain barrier and brain parenchyma 3, 15 , and creating more relevant architectures using threedimensional cultures 16-18. In addition, engineered platforms have been designed to enable controlled placement of neurons (e.g. cortical, hippocampal, amygdala) to characterize region-specific networks 19, 20 , or to isolate axons (or axonal bundles) for analysis 21, 22. Electrophysiological features of rodent-derived neural networks, established with both glutamatergic and GABAergic neurons, have been well characterized using dissociated neurons from primary cells or derived from neural stem cells 23-25. However, these systems most of...

RESUMOEste estudo teve por objetivo avaliar a freqüência de complicações materno-fetais, tipo de parto e controle metabólico das gestantes diabéticas atendidas no HCFMRP-USP, entre 1992 e 1999. Foram estudadas 261 pacientes, das quais 44 (16,3%) tinham diabetes mellitus tipo 1 (DM1), 82 (30,5%) diabetes tipo 2 (DM2) e 143 (53,2%) diabetes gestacional (DMG). Observou-se uma freqüência elevada de obesidade previamente à gestação nas pacientes com DMG (47,6%) e DM2 (65,9%) e também de HAS nesse último grupo (46,3%). Apesar do início tardio, houve no decorrer do seguimento melhora do controle metabólico do DM nos 3 grupos, observada através da redução dos níveis glicêmicos (DM1 às 10h: 197±40 vs. 128±39mg/dl, p= 0,003; DM2 às 7h: 147±53 vs. 102±19mg/dl, p= 0,001; 14h: 164±53 vs. 121±28mg/dl, p= 0,01; e 20h: 201±55 vs. 147±43mg/dl, p= 0,01; DMG às 7h: 100±34 vs. 89±20mg/dl, p= 0,003; 10h: 144±49 vs. 122±29mg/dl, p= 0,03 e 14h: 126±38 vs. 112±27mg/dl, p= 0,001) e da HbA1 (DM1: 11,1±2,9 vs. 5,7±1,8%, p= 0,02). As complicações maternas mais incidentes foram hipoglicemias, infecções do trato urinário (ITU), vulvovaginites, hipertensão arterial sistêmica (HAS) e doença hipertensiva específica da gravidez (DHEG). Foi significativamente maior a freqüência de hipoglicemias (29,5%, p< 0,0001), de ITU (29,5%, p= 0,02) e de abortos (11,4%, p= 0,003) nas pacientes com DM1 que nos outros grupos. Não houve nenhum óbito materno. Nos 3 grupos, o parto cesária foi o mais utilizado (DM1: 74,3%; DM2: 79,5%; DMG: 60,5%). Hipoglicemias, prematuridade, icterícia e macrossomia foram as complicações fetais de maior incidência. Foram complicações significativamente mais freqüentes nos recém-nascidos de gestantes com DM1: prematuridade (53,7%, p< 0,0001), natimortalidade (14,6%, p< 0,0001) e síndrome do desconforto respiratório do recém-nascido (13,9%, p= 0,003). Embora tenha havido melhora do controle metabólico nos grupos estudados, não foi atingida uma completa normalização dos níveis sangüíneos de glicose e hemoglobina glicada, o que provavelmente contribuiu para as taxas de complicações materno-fetais verificadas nas nossas pacientes. ABSTRACTThe objective of the present study was to evaluate the frequency of maternal and fetal complications, type of delivery and metabolic control of diabetic pregnant women followed at HCFMRP-USP, between 1992 and 1999. Outcome data were obtained on 261 patients where 44 (16.3%) had type 1 (DM1), 82 (30.5%) had type 2 (DM2) and 143 (53.2%) had gestational diabetes mellitus (GDM). The occurrence of obesity prior to gestation was elevated in patients with GDM (47.6%) and DM2 (65.9%). Hypertension was also frequent in the latter group (46.3%). In spite of the late beginning of the prenatal care in all three groups, there was metabolic control improvement, observed through reduction of glycemias (DM1: 10h: 197±40 vs. 128±39mg/dl, p= 0.003; DM2: 7h: 147±53 vs. 102±19mg/dl, p= 0.001; 14h: 164±53 vs. 121±28mg/dl, p= 0.01; 20h: 201±55 vs. 147±43mg/dl, p= 0.01; artigo originalEvolução Materno-Fetal de Gestantes ...

The widespread adoption of electronic medical records (EMRs) in healthcare has provided vast new amounts of data for statistical machine learning researchers in their efforts to model and predict patient health status, potentially enabling novel advances in treatment. In the case of sepsis, a debilitating, dysregulated host response to infection, extracting subtle, uncataloged clinical phenotypes from the EMR with statistical machine learning methods has the potential to impact patient diagnosis and treatment early in the course of their hospitalization. However, there are significant barriers that must be overcome to extract these insights from EMR data. First, EMR datasets consist of both static and dynamic observations of discrete and continuous-valued variables, many of which may be missing, precluding the application of standard multivariate analysis techniques. Second, clinical populations observed via EMRs and relevant to the study and management of conditions like sepsis are often heterogeneous; properly accounting for this heterogeneity is critical. Here, we describe an unsupervised, probabilistic framework called a composite mixture model that can simultaneously accommodate the wide variety of observations frequently observed in EMR datasets, characterize heterogeneous clinical populations, and handle missing observations. We demonstrate the efficacy of our approach on a large-scale sepsis cohort, developing novel techniques built on our model-based clusters to track patient mortality risk over time and identify physiological trends and distinct subgroups of the dataset associated with elevated risk of mortality during hospitalization.