Search

Objective Estimating the deterioration paths of chronic hepatitis B (CHB) patients is critical for physicians’ decisions and patient management. A novel, hierarchical multilabel graph attention-based method aims to predict patient deterioration paths more effectively. Applied to a CHB patient data set, it offers strong predictive utilities and clinical value. Materials and Methods The proposed method incorporates patients’ responses to medications, diagnosis event sequences, and outcome dependencies to estimate deterioration paths. From the electronic health records maintained by a major healthcare organization in Taiwan, we collect clinical data about 177 959 patients diagnosed with hepatitis B virus infection. We use this sample to evaluate the proposed method’s predictive efficacy relative to 9 existing methods, as measured by precision, recall, F-measure, and area under the curve (AUC). Results We use 20% of the sample as holdouts to test each method’s prediction performance. The results indicate that our method consistently and significantly outperforms all benchmark methods. It attains the highest AUC, with a 4.8% improvement over the best-performing benchmark, as well as 20.9% and 11.4% improvements in precision and F-measures, respectively. The comparative results demonstrate that our method is more effective for predicting CHB patients’ deterioration paths than existing predictive methods. Discussion and Conclusion The proposed method underscores the value of patient-medication interactions, temporal sequential patterns of distinct diagnosis, and patient outcome dependencies for capturing dynamics that underpin patient deterioration over time. Its efficacious estimates grant physicians a more holistic view of patient progressions and can enhance their clinical decision-making and patient management.

Multiplex immunofluorescence (mIF) assays multiple protein biomarkers on a single tissue section. Recently, high-plex CODEX (co-detection by indexing) systems enable simultaneous imaging of 40+ protein biomarkers, unlocking more detailed molecular phenotyping, leading to richer insights into cellular interactions and disease. However, high-plex data can be slower and more costly to collect, limiting its applications, especially in clinical settings. We propose a machine learning framework, 7-UP, that can computationally generate in silico 40-plex CODEX at single-cell resolution from a standard 7-plex mIF panel by leveraging cellular morphology. We demonstrate the usefulness of the imputed biomarkers in accurately classifying cell types and predicting patient survival outcomes. Furthermore, 7-UP’s imputations generalize well across samples from different clinical sites and cancer types. 7-UP opens the possibility of in silico CODEX, making insights from high-plex mIF more widely available.

Artificial Intelligence (AI) in healthcare has generated remarkable innovation and progress in the last decade. Significant advancements can be attributed to the utilization of AI to transform physiology data to advance healthcare. In this review, we will explore how past work has shaped the field and defined future challenges and directions. In particular, we focus on three areas of development. First, we give an overview of AI, with special attention to the most relevant AI models. We then detail how physiology data has been harnessed by AI to advance the main areas of healthcare such as automating existing healthcare tasks, increasing access to care, and augmenting healthcare capabilities. Finally, we discuss emerging concerns surrounding the use of individual physiology data and detail an increasingly important consideration for the field, namely the challenges of deploying AI models to achieve meaningful clinical impact.

Background This network meta-analysis was performed to determine the optimal surgical approach for pancreatoduodenectomy by comparing outcomes after laparoscopic pancreatoduodenectomy, robotic pancreatoduodenectomy and open pancreatoduodenectomy. Methods A systematic search of the PubMed, EMBASE, Scopus, and Web of Science databases was conducted to identify eligible randomized controlled trials and propensity-score matched studies. Results Four randomized controlled trials and 23 propensity-score matched studies comprising a total of 4,945 patients were included for analysis. Operation time for open pancreatoduodenectomy was shorter than both laparoscopic pancreatoduodenectomy (mean difference –57.35, 95% CI 26.25–88.46 minutes) and robotic pancreatoduodenectomy (mean difference –91.08, 95% CI 48.61–133.56 minutes), blood loss for robotic pancreatoduodenectomy was significantly less than both laparoscopic pancreatoduodenectomy (mean difference –112.58, 95% CI 36.95–118.20 mL) and open pancreatoduodenectomy (mean difference –209.87, 95% CI 140.39–279.36 mL), both robotic pancreatoduodenectomy and laparoscopic pancreatoduodenectomy were associated with reduced rates of delayed gastric emptying compared with open pancreatoduodenectomy (odds ratio 0.59, 95% CI 0.39–0.90 and odds ratio 0.69, 95% CI 0.50–0.95, respectively), robotic pancreatoduodenectomy was associated with fewer wound infections compared with open pancreatoduodenectomy (odds ratio 0.35, 95% CI 0.18–0.71), and laparoscopic pancreatoduodenectomy patients enjoyed significantly shorter length of stay compared with open pancreatoduodenectomy (odds ratio 0.43, 95% CI 0.28–0.95). There were no differences in other outcomes. Conclusion This network meta-analysis of high-quality studies suggests that when laparoscopic pancreatoduodenectomy and robotic pancreatoduodenectomy are performed in high-volume centers, short-term perioperative and oncologic outcomes are largely comparable, if not slightly improved, compared with traditional open pancreatoduodenectomy. These findings should be corroborated in further prospective randomized studies.

Background The exact role of laparoscopic liver resection (LLR) in patients with hepatocellular carcinoma (HCC) and underlying liver cirrhosis (LC) is not well defined. In this meta-analysis, both long- and short-term outcomes following LLR versus open liver resection (OLR) were analysed. Methods PubMed, EMBASE, Scopus and Web of Science databases were searched systematically for randomised controlled trials (RCTs) and propensity-score matched (PSM) studies reporting outcomes of LLR versus OLR of HCC in patients with cirrhosis. Primary outcome was overall survival (OS). This was analysed using one-stage (individual participant data meta-analysis) and two-stage (aggregate data meta-analysis) approaches. Secondary outcomes were operation duration, blood loss, blood transfusion, Pringle manoeuvre utilization, overall and major complications, length of hospital stay (LOHS), 90-day mortality and R0 resection rates. Results Eleven studies comprising 1618 patients (690 LLR versus 928 OLR) were included for analysis. In the one-stage meta-analysis, an approximately 18.7 per cent lower hazard rate (HR) of death in the LLR group (random effects: HR 0.81, 95 per cent confidence interval [C.I.] 0.68 to 0.96; P = 0.018) was observed. Two-stage meta-analysis resulted in a pooled HR of 0.84 (95 per cent C.I. 0.74 to 0.96; P = 0.01) in the overall LLR cohort. This indicated a 16–26 per cent reduction in the HR of death for patients with HCC and cirrhosis who underwent LLR. For secondary outcomes, LLR was associated with less blood loss (mean difference [MD] −99 ml, 95 per cent C.I. −182 to −16 ml), reduced overall complications (odds ratio 0.49, 95 per cent C.I. 0.37 to 0.66) and major complications (odds ratio 0.45, 95 per cent C.I. 0.26 to 0.79), and shorter LOHS (MD −3.22 days, 95 per cent C.I. −4.38 to −2.06 days). Conclusion Laparoscopic resection of HCC in patients with cirrhosis is associated with improved survival and perioperative outcomes.

In photocatalytic schemes, it is common practice to employ a sacrificial agent for scavenging photogenerated holes and enhancing the yield and efficiency of electron harvesting. However, little att...

Multiplex immunofluorescence (mIF) assays multiple protein biomarkers on a single tissue section. Recently, high-plex CODEX (co-detection by indexing) systems enable simultaneous imaging of 40+ protein biomarkers, unlocking more detailed molecular phenotyping, leading to richer insights into cellular interactions and disease. However, high-plex imaging can be slower and more costly to collect, limiting its applications, especially in clinical settings. We propose a machine learning framework, 7-UP, that can computationally generate in silico 40-plex CODEX at single-cell resolution from a standard 7-plex mIF panel by leveraging cellular morphology. We demonstrate the usefulness of the imputed biomarkers in accurately classifying cell types and predicting patient survival outcomes. Furthermore, 7-UP’s imputations generalize well across samples from different clinical sites and cancer types. 7-UP opens the possibility of in silico CODEX, making insights from high-plex mIF more widely available.

Visual information arriving at the retina is transmitted to the brain by signals in the optic nerve, and the brain must rely solely on these signals to make inferences about the visual world. Previous work has probed the content of these signals by directly reconstructing images from retinal activity using linear regression or nonlinear regression with neural networks. Maximuma posteriori(MAP) reconstruction using retinal encoding models and separately-trained natural image priors offers a more general and principled approach. We develop a novel method for approximate MAP reconstruction that combines a generalized linear model for retinal responses to light, including their dependence on spike history and spikes of neighboring cells, with the image prior implicitly embedded in a deep convolutional neural network trained for image denoising. We use this method to reconstruct natural images fromex vivosimultaneously-recorded spikes of hundreds of retinal ganglion cells uniformly sampling a region of the retina. The method produces reconstructions that match or exceed the state-of-the-art in perceptual similarity and exhibit additional fine detail, while using substantially fewer model parameters than previous approaches. The use of more rudimentary encoding models (a linear-nonlinear-Poisson cascade) or image priors (a 1/fspectral model) significantly reduces reconstruction performance, indicating the essential role of both components in achieving high-quality reconstructed images from the retinal signal.

Multiplexed immunofluorescence imaging allows the multidimensional molecular profiling of cellular environments at subcellular resolution. However, identifying and characterizing disease-relevant microenvironments from these rich datasets is challenging. Here we show that a graph neural network that leverages spatial protein profiles in tissue specimens to model tumour microenvironments as local subgraphs captures distinctive cellular interactions associated with differential clinical outcomes. We applied this spatial cellular-graph strategy to specimens of human head-and-neck and colorectal cancers assayed with 40-plex immunofluorescence imaging to identify spatial motifs associated with cancer recurrence and with patient survival after treatment. The graph deep learning model was substantially more accurate in predicting patient outcomes than deep learning approaches that model spatial data on the basis of the local composition of cell types, and it generated insights into the effect of the spatial compartmentalization of tumour cells and granulocytes on patient prognosis. Local graphs may also aid in the analysis of disease-relevant motifs in histology samples characterized via spatial transcriptomics and other -omics techniques.

Multiplexed immunofluorescence imaging enables high-dimensional molecular profiling at subcellular resolution. However, learning disease-relevant cellular environments from these rich imaging data is an open challenge. We developed SPAtial CEllular Graphical Modeling (SPACE-GM), a geometric deep learning framework that flexibly models tumor microenvironments (TMEs) as cellular graphs. We applied SPACE-GM to 658 head-and-neck and colorectal human cancer samples assayed with 40-plex immunofluorescence imaging to identify spatial motifs associated with cancer recurrence and patient survival after immunotherapy. SPACE-GM is substantially more accurate in predicting patient outcomes than previous approaches for modeling spatial data using neighborhood cell-type compositions. Computational interpretation of the disease-relevant microenvironments identified by SPACE-GM generates insights into the effect of spatial dispersion of tumor cells and granulocytes on patient prognosis.

Clinical trials are the gatekeepers and bottlenecks of progress in medicine. In recent years, they have become increasingly complex and expensive, driven by a growing number of stakeholders requiring more endpoints, more diverse patient populations, and a stringent regulatory environment. Trial designers have historically relied on investigator expertise and legacy norms established within sponsor companies to improve operational efficiency while achieving study goals. As such, data-driven forecasts of operational metrics can be a useful resource for trial design and planning. We develop a machine learning model to predict clinical trial operational efficiency using a novel dataset from Roche containing over 2,000 clinical trials across 20 years and multiple disease areas. The data includes important operational metrics related to patient recruitment and trial duration, as well as a variety of trial features such as the number of procedures, eligibility criteria, and endpoints. Our results demonstrate that operational efficiency can be predicted robustly using trial features, which can provide useful insights to trial designers on the potential impact of their decisions on patient recruitment success and trial duration.

Background: Despite advances in tyrosine kinase inhibitor (TKI) therapy for chronic myeloid leukemia in chronic phase (CML-CP), a sizeable proportion of patients with CML-CP remains refractory or intolerant to these agents. Objectives: Treatment patterns, healthcare resource utilization (HRU), and costs were evaluated among patients with CML who received third or later lines of therapy (3L+), a clinical population that has not been previously well-studied, with unmet treatment needs as TKI therapy has repeatedly failed. Methods: Adult patients with CML who received 3L+ were identified in the IBM® MarketScan® Databases (January 1, 2001–June 30, 2019) and the SEER-Medicare–linked database (January 1, 2006–December 31, 2016). Treatment patterns were observed from CML diagnosis. HRU and direct healthcare costs (payer’s perspective, 2019 USD) were measured in a 3L+ setting. Results: Among 296 commercially insured patients with 3L+ (median age, 58.5 years; female, 49.7%), the median duration of first-line (1L), second-line (2L), and 3L therapy was 8.5, 4.2, and 8.3 months, respectively. The annual incidence rate during 3L+ was 3.4 for inpatient days, 30.8 for days with outpatient services, and 1.2 for emergency department visits. Mean per-patient-per-month (PPPM) total healthcare costs (pharmacy + medical costs) were $18 784 in 3L+, $15 206 in 3L, and $19 546 in 4L, with inpatient costs driving most of the difference between 3L and 4L (mean [3L] = $2528 PPPM, mean [4L] = $6847 PPPM). Among 53 Medicare-insured patients with 3L+ (median age, 72.0 years; female, 39.6%), the median duration of 1L, 2L, and 3L therapy was 9.7, 5.0, and 7.0 months, respectively. During 3L+, the annual incidence rate was 10.3 for inpatient days, 61.9 for days with outpatient services, and 1.5 for emergency department visits. Mean PPPM total healthcare costs were $14 311 in 3L+, $15 100 in 3L, and $16 062 in 4L. Discussion: Patients with CML receiving 3L+ rapidly cycled through multiple lines. Costs increased from 3L to 4L; in commercially insured patients, inpatient costs were responsible for most of the cost increase between 3L and 4L, underlying these patients’ continued need for care. Conclusions: These findings support the need for better treatment options in patients with CML undergoing later lines of therapy.

Due to limited treatment options, many patients with diabetic gastroparesis (DG) or idiopathic gastroparesis (IG) experience inadequate symptom control resulting in increased health-care resource utilization (HRU) and associated costs. We compared all-cause HRU and health-care costs over the 3 years after patients' first gastroparesis diagnosis with that of matched controls without gastroparesis.Newly diagnosed adults with DG or IG were identified in Optum's de-identified ClinformaticsDemographics and comorbidities were balanced between patients with gastroparesis (n = 18,015 [DG]; n = 14,305 [IG]) and controls. In each of the Years 1-3 post-index, patients with DG or IG had significantly higher annual HRU and costs versus controls (mean total cost differences PPY: DG Year 1 $34,885, Year 2 $28,071, Year 3 $25,606; IG Year 1 $23,176, Year 2 $16,627, Year 3 $14,396) (all p 0.05). Across all 3 years, DG/IG cohorts had approximately twice the costs of controls. HRU and costs were highest in Year 1 post-index for both DG and IG.The economic burden of gastroparesis remains high several years after diagnosis, emphasizing the need for chronic treatment to effectively manage symptoms and consequently reduce the burden of this disorder.

Radical antegrade modular pancreatosplenectomy (RAMPS) was introduced to improve the tangential resection margin rates and N1 node clearance following resection of malignancies of the pancreatic body and tail. Owing to its technical complexity, minimally invasive RAMPS (MI-RAMPS) has only been reported by a few centers worldwide. We performed this meta-analysis to compare both short- and long-term outcomes between open RAMPS (O-RAMPS) and minimally invasive RAMPS (MI-RAMPS). A systematic search of the electronic databases PubMed, Medline (via PubMed), Cochrane Register of Controlled Trials (CENTRAL), EMBASE, Scopus and Web of Science was performed to identify eligible studies published in the English language regardless of study design. The outcomes of interest were operation time, estimated blood loss, transfusion rates, overall complications, Grade B/C post-operative pancreatic fistula (POPF) rates, post-pancreatectomy hemorrhage (PPH), delayed gastric emptying (DGE), length of stay (LOS), R0 resection rates, lymph node (LN) yield and overall survival (OS). Five non-randomized studies comprising of a total 229 patients (89 MI-RAMPS, 140 O-RAMPS) were included for analysis. Intra-operative blood loss was observed to be significantly reduced in MI-RAMPS as compared to O-RAMPS (MD –256.16, P

Etavopivat (FT-4202) is an orally administered, small-molecule allosteric activator of erythrocyte pyruvate kinase-R (PKR) in clinical development for the treatment of sickle cell disease and other hemoglobin disorders. This randomized, placebo-controlled, double-blind, first-in-human combination single-ascending dose and multiple-ascending dose phase 1 trial (NCT03815695) evaluated the safety and pharmacokinetics/pharmacodynamics of etavopivat in 90 healthy adult subjects. In 4 single-ascending dose cohorts, 8 participants were randomized 3:1 to a single oral dose of either etavopivat (n = 6) or placebo (n = 2). In four 14-day multiple-ascending dose cohorts, 12 participants were randomized 3:1 to 14 days of etavopivat (n = 9) or placebo (n = 3). In these studies, most treatment-emergent adverse events were of mild severity (grade 1) and none led to study discontinuation. Etavopivat exhibited a linear and time-independent pharmacokinetic profile (at doses ≤400 mg) and elicited the expected pharmacodynamic effects of PKR activation (decreased 2,3-diphosphoglycerate and increased adenosine triphosphate) and evidence of improved hemoglobin-oxygen affinity. In addition, pharmacodynamic responses were durable with effects continuing for 48 to 72 hours after the last dose, thereby supporting once-daily dosing. Food appeared to have no clinically meaningful effects on etavopivat exposure, thus facilitating administration with or without food. In conclusion, the evaluation of etavopivat in healthy subjects demonstrated proof of mechanism (PKR activation) without significant adverse events. This study also allowed for the selection of dose levels, projected to have an acceptable safety profile and provide therapeutic benefit, for evaluation in future trials in patients with sickle cell disease.

Various real-world data (RWD) sources have emerged in China with the intention of generating real-world evidence (RWE) that can be used in clinical and regulatory decision-making. Despite these efforts, significant barriers remain that hinder high-quality healthcare research. A workshop with 30 representatives from healthcare research agencies, technology companies focused on healthcare big data and pharmaceutical companies was held in December 2020 to identify strategies to overcome the barriers associated with the usability and quality of RWD in China. Across all sectors, examples of barriers identified included inconsistencies in terminology and non-standardised coding practices; the absence of longitudinal data; the absence of transparent data processing and validation practices; and the inability to access and share RWD. While cutting-edge technological innovations and data solutions provided powerful tools, the development of collaborative and synergistic research networks across multiple stakeholders is key to generate accessible, high-quality RWD in China. RWD has the potential to provide clinical, regulatory and reimbursement decision-makers with critical insights that can improve healthcare delivery in China. However, barriers to its access, collection and use must be addressed to generate RWE to guide healthcare stakeholders.

A comprehensive overview of medical AI devices approved by the US Food and Drug Administration sheds new light on limitations of the evaluation process that can mask vulnerabilities of devices when they are deployed on patients.

Bunsen reaction ( 2 H 2 O + I 2 + S O 2 → H 2 S O 4 + 2 H I ) is a key step for hydrogen production from either the H2S splitting cycle or the sulfur-iodine (S-I) cycle of water splitting. As pointed out in part one, when engineering this reaction, many challenges such as side reactions and corrosion impede scaling up this process. Using iodine-toluene solution to provide flowing iodine below the melting point of iodine renders the Bunsen reaction to be conducted at ambient temperature such that these challenges can be either overcome or eased. However, using toluene as the iodine solvent makes the Bunsen reaction a multiphase reaction system which includes gas, aqueous, and organic phases. Glass-made Corning® advanced-flowTM reactors (AFRs) can be used for Bunsen reaction because they are good at resisting corrosion, improving mixing efficiency of multiphase fluids, and allowing seamless scaling up. Part one has studied the absorption behavior of SO2 gas in the liquids used for Bunsen reaction (water, toluene and water-toluene mixture). Part two (this work) mainly studies the Bunsen reaction using the Corning® microscale (LF) and milliscale (G1) AFRs. When I2 was dissolved in toluene, the Bunsen reaction was conducted by feeding SO2 gas, water, and I2/toluene solution into the AFRs. SO2 and I2 were used as the limiting reactants in turn, and the effects of operating conditions such as gas and liquid flow rates, water to toluene ratio, and temperature in the range (22-80 oC) on the absorption rates of SO2 and the I2 reaction rate were studied. The results confirm the seamless scaling-up capability of the Corning reactors when the flow rates were increased twenty times from AFR-LF to AFR-G1.

This study examines the quality of outpatient preventive care in a large, ethnically diverse cohort of outpatients with cirrhosis at a high‐volume tertiary liver transplant center. Receipt of evidence‐based quality indicators, including preventive processes for ascites, esophageal varices, hepatic encephalopathy, hepatocellular carcinoma, liver transplantation, and general cirrhosis care, was variable and overall low. These findings suggest that larger system redesign efforts are needed to improve the quality of care, beyond just improving access to specialty providers., We sought to identify specific gaps in preventive care provided to outpatients with cirrhosis and to determine factors associated with high quality of care (QOC), to guide quality improvement efforts. Outpatients with cirrhosis who received care at a large, academic tertiary health care system in the United States were included. Twelve quality indicators (QIs), including preventive care processes for ascites, esophageal varices, hepatic encephalopathy, hepatocellular carcinoma (HCC), and general cirrhosis care, were measured. QI pass rates were calculated as the proportion of patients eligible for a QI who received that QI during the study period. We performed logistic regression to determine predictors of high QOC (≥ 75% of eligible QIs) and receipt of HCC surveillance. Of the 439 patients, the median age was 63 years, 59% were male, and 19% were Hispanic. The median Model for End‐Stage Liver Disease–Sodium score was 11, 64% were compensated, and 32% had hepatitis C virus. QI pass rates varied by individual QIs, but were overall low. For example, 24% received appropriate HCC surveillance, 32% received an index endoscopy for varices screening, and 21% received secondary prophylaxis for spontaneous bacterial peritonitis. In multivariable analyses, Asian race (odds ratio [OR]: 3.7, 95% confidence interval [CI]: 1.3‐10.2) was associated with higher QOC, and both Asian race (OR: 3.3, 95% CI: 1.2‐9.0) and decompensated status (OR: 2.1, 95% CI: 1.1‐4.2) were associated with receipt of HCC surveillance. A greater number of specialty care visits was not associated with higher QOC. Conclusion: Receipt of outpatient preventive cirrhosis QIs was variable and overall low in a diverse cohort of patients with cirrhosis. Variation in care by race/ethnicity and illness trajectory should prompt further inquiry into identifying modifiable factors to standardize care delivery and to improve QOC.

Mast cells are potent mediators of allergy and asthma, yet their role in regulating adaptive immunity remains ambiguous. On the surface of mast cells, the crosslinking of IgE bound to FcεRI by a specific antigen recognized by that IgE triggers the release of immune mediators such as histamine and cytokines capable of activating other immune cells; however, little is known about the mast cell contribution to the induction of endogenous, antigen-specific CD4+ T cells. Here we examined the effects of specific mast cell activation in vivo on the initiation of an antigen-specific CD4+ T cell response. While CD4+ T cells were not enhanced by FcεRI stimulation alone, their activation was synergistically enhanced when FcεRI activation was combined with TLR4 stimulation. This enhanced activation was dependent on global TLR4 stimulation but appeared to be less dependent on mast cell expressed TLR4. This study provides important new evidence to support the role of mast cells as mediators of the antigen-specific adaptive immune response.

Electrical stimulation of retinal ganglion cells (RGCs), which transmit visual information to the brain, is used in retinal implants to treat blindness caused by photoreceptor degeneration. However, the performance of existing clinical implants is limited by indiscriminate stimulation of many cells and cell types. Recent work in isolated macaque retina has shown the ability to precisely evoke spikes in the major RGC types by direct electrical stimulation at safe current levels, with single-cell, single-spike resolution and avoidance of axon bundle activation in many cases. However, these findings have not been verified in the human retina. Here, electrical activation of the major human RGC types was examined using large-scale, multi-electrode recording and stimulation and compared to results from several macaque retinas obtained using the same methods. Electrical stimulation of the major human RGC types closely paralleled results in macaque, with similar somatic and axonal stimulation thresholds, cellular and cell type selectivity of stimulation, avoidance of axon bundle stimulation by calibration, targeting of different cell types based on their distinct electrical signatures, and potential efficacy of real-time stimulus optimization for artificial vision. The results indicate that the macaque retina provides a quantitatively accurate picture of how focal electrical stimulation can be used in future high-resolution implants.

The functional organization of diverse retinal ganglion cell (RGC) types, which shapes the visual signal transmitted to the brain, has been examined in many species. The unique spatial, temporal, and chromatic properties of the numerically dominant RGC types in macaque monkey retina are presumed to most accurately model human vision. However, the functional similarity between RGCs in macaques and humans has only begun to be tested, and recent work suggests possible differences. Here, the properties of the numerically dominant human RGC types were examined using large-scale multi-electrode recordings with fine-grained visual stimulation in isolated retina, and compared to results from dozens of recordings from macaque retina using the same experimental methods and conditions. The properties of four major human RGC types -- ON-parasol, OFF-parasol, ON-midget, and OFF-midget -- closely paralleled those of the same macaque RGC types, including the spatial and temporal light sensitivity, precisely coordinated mosaic organization of receptive fields, ON-OFF asymmetries, spatial response nonlinearity, and sampling of photoreceptor inputs over space. Putative smooth monostratified cells and polyaxonal amacrine cells were also identified based on similarities to cell types previously identified in macaque retina. The results suggest that recently proposed differences between human and macaque RGCs probably reflect experimental differences, and that the macaque model provides an accurate picture of human RGC function.

To examine the extent to which there is a basic economic case for the provision of rehabilitation services within a publicly funded, device-centred approach to adult hearing services.Five representative cross-sectional surveys of people aged 50 years and over were conducted in Australia between 2013 and 2019 (Among respondents 27% reported that their hearing was fair or poor, and 14% reported hearing aid ownership. 45% of hearing aid owners were regular device users. Approximately 1:4 people (24%) who owned a hearing aid had not used it in the past three months; with an estimated cost of $au87.4 million per annum. The provision of hearing rehabilitation services in addition to or instead of hearing aids could provide clients with a more comprehensive service at an estimated saving of between 62% to 81% of existing program costs.Cost benefit analysis supports the provision of hearing rehabilitation within a hearing services program, either as an alternative to the existing service, or adjunct to it. Such a service would enhance client outcomes and reduce the level of wastage currently associated with the current device-centred approach.Implications for rehabilitationClients identified as not being immediately ready to progress to hearing aids could benefit from rehabilitation programs while still having access to either assistive listening devices or hearing aids.The proposed approach could result in significant cost savings to publicly funded programs while ensuring better outcomes for service users through the more effective, efficient and ethical use of public monies.

Breast cancer remains a global challenge, causing over 600,000 deaths in 2018 (ref. 1). To achieve earlier cancer detection, health organizations worldwide recommend screening mammography, which is estimated to decrease breast cancer mortality by 20–40% (refs. 2,3). Despite the clear value of screening mammography, significant false positive and false negative rates along with non-uniformities in expert reader availability leave opportunities for improving quality and access4,5. To address these limitations, there has been much recent interest in applying deep learning to mammography6–18, and these efforts have highlighted two key difficulties: obtaining large amounts of annotated training data and ensuring generalization across populations, acquisition equipment and modalities. Here we present an annotation-efficient deep learning approach that (1) achieves state-of-the-art performance in mammogram classification, (2) successfully extends to digital breast tomosynthesis (DBT; ‘3D mammography’), (3) detects cancers in clinically negative prior mammograms of patients with cancer, (4) generalizes well to a population with low screening rates and (5) outperforms five out of five full-time breast-imaging specialists with an average increase in sensitivity of 14%. By creating new ‘maximum suspicion projection’ (MSP) images from DBT data, our progressively trained, multiple-instance learning approach effectively trains on DBT exams using only breast-level labels while maintaining localization-based interpretability. Altogether, our results demonstrate promise towards software that can improve the accuracy of and access to screening mammography worldwide. A generalizable and interpretable artificial-intelligence system achieves clinical accuracy for screening and early breast-cancer detection on 2D and 3D mammograms.

Background An estimated 6500 undocumented immigrants with end-stage renal disease (ESRD) live in the United States. Those living in states that do not provide undocumented immigrants scheduled hemodialysis receive intermittent hemodialysis only when life-threatening conditions arise. Little is known about catheter-related bloodstream infections (CRBSIs) in this population. Methods We conducted a retrospective cohort study of emergency-only hemodialysis patients in the Harris Health System in Houston, Texas, between January 2012 and December 2015. We assessed CRBSI risk factors including demographics, comorbidities, and duration and frequency of hemodialysis. We investigated the microbiologic etiology of these infections, rates of recurrent CRBSI, and associated morbidity and mortality. Results The cohort included 329 patients; 90% were Hispanic, 60% had diabetes, and the average age was 51 years. A total of 101 CRBSIs occurred, with a rate of 0.84 infections per 1000 catheter-days. Cirrhosis and duration of hemodialysis during the study period were associated with increased risk of CRBSI. Seventeen CRBSIs were recurrent; infection with gram-positive bacteria predicted recurrence. Adherence to catheter-related infection guidelines was improved by infectious diseases consultation and associated with fewer recurrent infections. CRBSI was associated with prolonged hospitalization (mean, 15 days), composite complication rate of 8%, and a 4% mortality rate. Conclusions Patients receiving emergency-only hemodialysis via tunneled catheters have a high CRBSI rate compared with infection rates previously reported in patients receiving scheduled maintenance hemodialysis. Increased CRSBI risk likely contributes to the increased morbidity and mortality seen in ESRD patients receiving emergency-only hemodialysis.

This article aims to demonstrate how innovative testing practices can effectively prevent high-tech mass cheating and improve fairness in language assessment. The article first introduces Xi's (2010) view of validity and fairness and her proposal of an argument-based approach to empirically examining test fairness. The article then describes the threat to fair testing posed by high-tech cheating on the College English Test (CET). A study of multiple-form equating was conducted and reported in the article, which was aimed at achieving alternate form reliability when multiple versions and multiple forms were used in the CET. The article then concludes with a discussion on the usefulness of an argument-based approach to empirically examining test fairness.

This paper describes a post-race analysis of team KOMMIT EVT's electric motorcycle data collected during the 2016 Pikes Peak International Hill Climb (PPIHC). The motorcycle consumed approximately ...

We compared the performance of different Deep Learning - Convolutional Neural Network (DL-CNN) models for bladder cancer treatment response assessment based on transfer learning by freezing different DL-CNN layers and variation of the DL-CNN structure. Pre- and post-treatment CT scans of 123 patients (129 cancers, 158 pre- and posttreatment cancer pairs) undergoing chemotherapy were collected. 33% of patients had T0 stage cancer (complete response) after chemotherapy. Regions of interest (ROIs) of pre- and post-treatment scans were extracted from the segmented lesions and combined into hybrid pre-post image pairs. The dataset was split into training (94 pairs and 6209 hybrid ROIs), validation (10 pairs) and test sets (54 pairs). The DL-CNN consists of 2 convolution (C1, C2), 2 locally connected (L1, L2), and 1 fully connected layers, implemented in TensorFlow. The DL-CNN was trained to classify the bladder cancers as fully responding (stage T0) or not fully responding to chemotherapy based on the hybrid ROIs. Two blinded radiologists provided an estimate of the likelihood of the lesion being stage T0 post-treatment by reading the pairs of pre- and post-treatment CT volumes. The test AUC was 0.73 for T0 prediction by the base DL-CNN structure with randomly initialized weights. The base DL-CNN structure with transfer learning pre-trained weights (no frozen layers) achieved a test AUC of 0.79. The test AUCs for 3 modified DL-CNN structures (different C1, C2 max pooling filter sizes, strides, and padding, with transfer learning) were 0.72, 0.86, and 0.69, respectively. For the base DL-CNN with (C1) frozen, (C1, C2) frozen, and (C1, C2, L3) frozen during transfer learning, the test AUCs were 0.81, 0.78, and 0.71, respectively. The radiologists’ AUCs were 0.76 and 0.77. The DL-CNN performed better with pre-trained than randomly initialized weights.

We compared the performance of different Deep learning-convolutional neural network (DL-CNN) models for bladder cancer treatment response assessment based on transfer learning by freezing different DL-CNN layers and varying the DL-CNN structure. Pre- and posttreatment computed tomography scans of 123 patients (cancers, 129, pre- and posttreatment cancer pairs, 158) undergoing chemotherapy were collected. After chemotherapy 33% of patients had T0 stage cancer (complete response). Regions of interest in pre- and posttreatment scans were extracted from the segmented lesions and combined into hybrid pre -post image pairs (h-ROIs). Training (pairs, 94, h-ROIs, 6209), validation (10 pairs) and test sets (54 pairs) were obtained. The DL-CNN consisted of 2 convolution (C1-C2), 2 locally connected (L3-L4), and 1 fully connected layers. The DL-CNN was trained with h-ROIs to classify cancers as fully responding (stage T0) or not fully responding to chemotherapy. Two radiologists provided lesion likelihood of being stage T0 posttreatment. The test area under the ROC curve (AUC) was 0.73 for T0 prediction by the base DL-CNN structure with randomly initialized weights. The base DL-CNN structure with pretrained weights and transfer learning (no frozen layers) achieved test AUC of 0.79. The test AUCs for 3 modified DL-CNN structures (different C1-C2 max pooling filter sizes, strides, and padding, with transfer learning) were 0.72, 0.86, and 0.69. For the base DL-CNN with (C1) frozen, (C1-C2) frozen, and (C1-C2-L3) frozen, the test AUCs were 0.81, 0.78, and 0.71, respectively. The radiologists' AUCs were 0.76 and 0.77. DL-CNN performed better with pretrained than randomly initialized weights.

Supplemental Digital Content is available in the text., Objectives: In patients with septic shock, hydrocortisone 200–400 mg/d has been shown to reverse shock compared with placebo. Lower doses of hydrocortisone have not previously been studied, and there are no previous studies comparing two different doses of hydrocortisone. At our institution, some clinicians routinely prescribe doses less than 200 mg/d. This study aims to compare the effect of lower doses of hydrocortisone to standard doses on shock reversal and adverse events in septic shock. Design: Retrospective cohort study. Setting: Single-center medical ICU. Subjects: Patients who received hydrocortisone for septic shock. Interventions: Electronic chart review. Measurements and Main Results: Patients were divided into low-dose hydrocortisone (75–150 mg/d) and standard-dose hydrocortisone (200–400 mg/d) cohorts based on initial prescribed hydrocortisone dose. Rates of shock reversal and adverse events in the two cohorts were compared. Two-hundred thirteen patients were included—41 in low-dose and 172 in standard-dose cohorts. Baseline characteristics including initial vasopressor requirement and Sequential Organ Failure Assessment scores were similar. Average rates of change in vasopressor needs, conditional hazard rate for vasopressor withdrawal, and cumulative probability for vasopressor withdrawal were all quantitatively similar for low-dose and standard-dose hydrocortisone. Insulin requirement (particularly in those with diabetes mellitus), blood glucose in those with diabetes mellitus, and frequency of secondary infections seemed to be lower in the low-dose hydrocortisone cohort. Mortality and other secondary outcomes were similar. Conclusions: In septic shock, hydrocortisone dosed 75–150 mg/d appears to reverse shock as effectively 200–400 mg/d and may cause a lower frequency of adverse events.

Bridge to recovery with left ventricular assist device (LVAD) support has been more prominent with volume displacement pumps (VDPs) than with rotary blood pumps (RBPs), which may be due to VDPs providing greater ventricular unloading and coronary artery flow. To compare ventricular unloading and coronary flow of VDPs and RBPs in a repeatable environment, a physiologic coronary circulation was added to a pre-existing mock circulatory loop. In this study, a physiologic coronary circulation, mimicking a healthy or diseased auto-regulatory response was implemented in a mock circulatory loop. Using the mock circulation loop, a VDP with original (Björk-Shiley) and then replacement (jellyfish) valves was operated in clinically recommended modes and compared to full and partial assist RBP operating at constant speed and rapid speed modulated modes. The Björk-Shiley VDP resulted in increased pressure-volume area, which resulted in greater coronary artery flow when compared to the improved jellyfish valves. Full assist RBP support reduced left ventricular stroke work, pressure-volume area and coronary flow compared to partial assist, whilst the effect of speed modulation modes was not as significant. Of all LVAD operating modes, the counter-pulsed VDP with jellyfish valves demonstrated the greatest reduction in pressure-volume area and improved coronary flow. This study provides a basis for further investigation into RBP speed modulation profiles to match the improved haemodynamic performance of VDPs.

Deep learning approaches to breast cancer detection in mammograms have recently shown promising results. However, such models are constrained by the limited size of publicly available mammography datasets, in large part due to privacy concerns and the high cost of generating expert annotations. Limited dataset size is further exacerbated by substantial class imbalance since “normal” images dramatically outnumber those with findings. Given the rapid progress of generative models in synthesizing realistic images, and the known effectiveness of simple data augmentation techniques (e.g. horizontal flipping), we ask if it is possible to synthetically augment mammogram datasets using generative adversarial networks (GANs). We train a class-conditional GAN to perform contextual in-filling, which we then use to synthesize lesions onto healthy screening mammograms. First, we show that GANs are capable of generating high-resolution synthetic mammogram patches. Next, we experimentally evaluate using the augmented dataset to improve breast cancer classification performance. We observe that a ResNet-50 classifier trained with GAN-augmented training data produces a higher AUROC compared to the same model trained only on traditionally augmented data, demonstrating the potential of our approach.