Deep learning offers considerable promise for medical diagnostics. We aimed to evaluate the diagnostic accuracy of deep learning algorithms versus health-care professionals in classifying diseases ...using medical imaging.
In this systematic review and meta-analysis, we searched Ovid-MEDLINE, Embase, Science Citation Index, and Conference Proceedings Citation Index for studies published from Jan 1, 2012, to June 6, 2019. Studies comparing the diagnostic performance of deep learning models and health-care professionals based on medical imaging, for any disease, were included. We excluded studies that used medical waveform data graphics material or investigated the accuracy of image segmentation rather than disease classification. We extracted binary diagnostic accuracy data and constructed contingency tables to derive the outcomes of interest: sensitivity and specificity. Studies undertaking an out-of-sample external validation were included in a meta-analysis, using a unified hierarchical model. This study is registered with PROSPERO, CRD42018091176.
Our search identified 31 587 studies, of which 82 (describing 147 patient cohorts) were included. 69 studies provided enough data to construct contingency tables, enabling calculation of test accuracy, with sensitivity ranging from 9·7% to 100·0% (mean 79·1%, SD 0·2) and specificity ranging from 38·9% to 100·0% (mean 88·3%, SD 0·1). An out-of-sample external validation was done in 25 studies, of which 14 made the comparison between deep learning models and health-care professionals in the same sample. Comparison of the performance between health-care professionals in these 14 studies, when restricting the analysis to the contingency table for each study reporting the highest accuracy, found a pooled sensitivity of 87·0% (95% CI 83·0-90·2) for deep learning models and 86·4% (79·9-91·0) for health-care professionals, and a pooled specificity of 92·5% (95% CI 85·1-96·4) for deep learning models and 90·5% (80·6-95·7) for health-care professionals.
Our review found the diagnostic performance of deep learning models to be equivalent to that of health-care professionals. However, a major finding of the review is that few studies presented externally validated results or compared the performance of deep learning models and health-care professionals using the same sample. Additionally, poor reporting is prevalent in deep learning studies, which limits reliable interpretation of the reported diagnostic accuracy. New reporting standards that address specific challenges of deep learning could improve future studies, enabling greater confidence in the results of future evaluations of this promising technology.
None.
Clear-cell carcinomas (CCCs) are a histological group of highly aggressive malignancies commonly originating in the kidney and ovary. CCCs are distinguished by aberrant lipid and glycogen ...accumulation and are refractory to a broad range of anti-cancer therapies. Here we identify an intrinsic vulnerability to ferroptosis associated with the unique metabolic state in CCCs. This vulnerability transcends lineage and genetic landscape, and can be exploited by inhibiting glutathione peroxidase 4 (GPX4) with small-molecules. Using CRISPR screening and lipidomic profiling, we identify the hypoxia-inducible factor (HIF) pathway as a driver of this vulnerability. In renal CCCs, HIF-2α selectively enriches polyunsaturated lipids, the rate-limiting substrates for lipid peroxidation, by activating the expression of hypoxia-inducible, lipid droplet-associated protein (HILPDA). Our study suggests targeting GPX4 as a therapeutic opportunity in CCCs, and highlights that therapeutic approaches can be identified on the basis of cell states manifested by morphological and metabolic features in hard-to-treat cancers.
Gilbert’s syndrome (GS) is characterized by a benign, mildly elevated bilirubin concentration in the blood. Recent reports show clear protection from cardiovascular disease in this population. ...Protection of lipids, proteins and other macromolecules from oxidation by bilirubin represents the most commonly accepted mechanism contributing to protection in this group. However, a recent meta-analysis estimated that bilirubin only accounts for ∼34% of the cardioprotective effects within analysed studies. To reveal the additional contributing variables we have explored circulating cholesterol and triacylglycerol concentrations, which appear to be decreased in hyperbilirubinemic individuals/animals, and are accompanied by lower body mass index in highly powered studies. These results suggest that bilirubin could be responsible for the development of a lean and hypolipidemic state in GS. Here we also discuss the possible contributing mechanisms that might reduce circulating cholesterol and triacylglycerol concentrations in individuals with syndromes affecting bilirubin metabolism/excretion, which we hope will stimulate future research in the area. In summary, this article is the first review of lipid status in animal and human studies of hyperbilirubinemia and explores possible mechanisms that could contribute to lowering circulating lipid parameters and further explain cardiovascular protection in Gilbert’s syndrome.
The aim of this study was to analyze the changing patterns of Listeria monocytogenes contamination in a cheese processing facility manufacturing a wide range of ready-to-eat products. ...Characterization of L. monocytogenes isolates included genotyping by pulsed-field gel electrophoresis (PFGE) and multi-locus sequence typing (MLST). Disinfectant-susceptibility tests and the assessment of L. monocytogenes survival in fresh cheese were also conducted. During the sampling period between 2010 and 2013, a total of 1284 environmental samples were investigated. Overall occurrence rates of Listeria spp. and L. monocytogenes were 21.9% and 19.5%, respectively. Identical L. monocytogenes genotypes were found in the food processing environment (FPE), raw materials and in products. Interventions after the sampling events changed contamination scenarios substantially. The high diversity of globally, widely distributed L. monocytogenes genotypes was reduced by identifying the major sources of contamination. Although susceptible to a broad range of disinfectants and cleaners, one dominant L. monocytogenes sequence type (ST) 5 could not be eradicated from drains and floors. Significantly, intense humidity and steam could be observed in all rooms and water residues were visible on floors due to increased cleaning strategies. This could explain the high L. monocytogenes contamination of the FPE (drains, shoes and floors) throughout the study (15.8%). The outcome of a challenge experiment in fresh cheese showed that L. monocytogenes could survive after 14days of storage at insufficient cooling temperatures (8 and 16°C). All efforts to reduce L. monocytogenes environmental contamination eventually led to a transition from dynamic to stable contamination scenarios. Consequently, implementation of systematic environmental monitoring via in-house systems should either aim for total avoidance of FPE colonization, or emphasize a first reduction of L. monocytogenes to sites where contamination of the processed product is unlikely. Drying of surfaces after cleaning is highly recommended to facilitate the L. monocytogenes eradication.
•The occurrence rate of L. monocytogenes in a contaminated cheese plant was 19.5%.•The high diversity of L. monocytogenes genotypes was reduced after interventions.•One dominant sequence type (ST) 5 could not be eradicated from drains and floors.•Systematic L. monocytogenes environmental monitoring improves intervention strategies.
Target-identification and mechanism-of-action studies have important roles in small-molecule probe and drug discovery. Biological and technological advances have resulted in the increasing use of ...cell-based assays to discover new biologically active small molecules. Such studies allow small-molecule action to be tested in a more disease-relevant setting at the outset, but they require follow-up studies to determine the precise protein target or targets responsible for the observed phenotype. Target identification can be approached by direct biochemical methods, genetic interactions or computational inference. In many cases, however, combinations of approaches may be required to fully characterize on-target and off-target effects and to understand mechanisms of small-molecule action.
The enthusiasm for phenotypic screening as an approach for small-molecule discovery has increased dramatically over the last several years. The recent increase in phenotype-based discoveries is in ...part due to advancements in phenotypic readouts in improved disease models that recapitulate clinically relevant biology in cell culture. Of course, a major historical barrier to using phenotypic assays in chemical biology has been the challenge in determining the mechanism of action (MoA) for compounds of interest. With the combination of medically inspired phenotypic screening and the development of modern MoA methods, we can now start implementing this approach in chemical probe and drug discovery. In this Perspective, we highlight recent advances in phenotypic readouts and MoA determination by discussing several case studies in which both activities were required for understanding the chemical biology involved and, in some cases, advancing toward clinical development.
•Phenotypic screening enables medically relevant small-molecule discovery•Modern mechanism-of-action methods fuel phenotypic approaches•Integrating these activities will help discover new chemistry and biology
Phenotypic screening has been a powerful approach for small-molecule discovery. However, determining the mechanism of action has been a major barrier to success. Here, we discuss several case studies that take a modern approach to both aspects, and identify new probes and clinical lead candidates.
Musculoskeletal providers are increasingly recognizing the importance of social factors and their association with health outcomes as they aim to develop more comprehensive models of care delivery. ...Such factors may account for some of the unexplained variation between pathophysiology and level of pain intensity and incapability experienced by people with common conditions, such as persistent nontraumatic knee pain secondary to osteoarthritis (OA). Although the association of one's social position (for example, income, employment, or education) with levels of pain and capability are often assessed in OA research, the relationship between aspects of social context (or unmet social needs) and such symptomatic and functional outcomes in persistent knee pain are less clear.
(1) Are unmet social needs associated with the level of capability in patients experiencing persistently painful nontraumatic knee conditions, accounting for sociodemographic factors? (2) Do unmet health-related social needs correlate with self-reported quality of life?
We performed a prospective, cross-sectional study between January 2021 and August 2021 at a university academic medical center providing comprehensive care for patients with persistent lower extremity joint pain secondary to nontraumatic conditions such as age-related knee OA. A final 125 patients were included (mean age 62 ± 10 years, 65% 81 of 125 women, 47% 59 of 125 identifying as White race, 36% 45 of 125 as Hispanic or Latino, and 48% 60 of 125 with safety-net insurance or Medicaid). We measured patient-reported outcomes of knee capability (Knee injury and Osteoarthritis Outcome Score for Joint Replacement), quality of life (Patient-Reported Outcome Measure Information System PROMIS Global Physical Health and PROMIS Global Mental Health), and unmet social needs (Accountable Health Communities Health-Related Social Needs Survey, accounting for insufficiencies related to housing, food, transportation, utilities, and interpersonal violence), as well as demographic factors.
After controlling for demographic factors such as insurance status, education attained, and household income, we found that reduced knee-specific capability was moderately associated with experiencing unmet social needs (including food insecurity, housing instability, transportation needs, utility needs, or interpersonal safety) (standardized beta regression coefficient β = -4.8 95% confidence interval -7.9 to -1.7; p = 0.002 and substantially associated with unemployment (β = -13 95% CI -23 to -3.8; p = 0.006); better knee-specific capability was substantially associated with having Medicare insurance (β = 12 95% CI 0.78 to 23; p = 0.04). After accounting for factors such as insurance status, education attained, and household income, we found that older age was associated with better general mental health (β = 0.20 95% CI 0.0031 to 0.39; p = 0.047) and with better physical health (β = 0.004 95% CI 0.0001 to 0.008; p = 0.04), but effect sizes were small to negligible, respectively.
There is an association of unmet social needs with level of capability and unemployment in patients with persistent nontraumatic knee pain. This finding signals a need for comprehensive care delivery for patients with persistent knee pain that screens for and responds to potentially modifiable social risk factors, including those based on one's social circumstances and context, to achieve better outcomes.
Level II, prognostic study.
Although the function of the mammalian pancreas hinges on complex interactions of distinct cell types, gene expression profiles have primarily been described with bulk mixtures. Here we implemented a ...droplet-based, single-cell RNA-seq method to determine the transcriptomes of over 12,000 individual pancreatic cells from four human donors and two mouse strains. Cells could be divided into 15 clusters that matched previously characterized cell types: all endocrine cell types, including rare epsilon-cells; exocrine cell types; vascular cells; Schwann cells; quiescent and activated stellate cells; and four types of immune cells. We detected subpopulations of ductal cells with distinct expression profiles and validated their existence with immuno-histochemistry stains. Moreover, among human beta- cells, we detected heterogeneity in the regulation of genes relating to functional maturation and levels of ER stress. Finally, we deconvolved bulk gene expression samples using the single-cell data to detect disease-associated differential expression. Our dataset provides a resource for the discovery of novel cell type-specific transcription factors, signaling receptors, and medically relevant genes.
Display omitted
•We report over 12,000 individual pancreatic cell transcriptomes in human and mouse•We detect novel expression of TFs, signaling receptors, and medically relevant genes•We identify subpopulations and heterogeneity within pancreatic cell types•We deconvolve bulk gene expression samples using the single-cell data
Single-cell transcriptomics of over 12,000 cells from four human donors and two mouse strains was determined using inDrop. Cells were divided into 15 clusters that matched previously characterized cell types. Detailed analysis of each population separately revealed subpopulations within the ductal population, modes of activation of stellate cells, and heterogeneity in the stress among beta cells.
A new PCCP‐coordinated molybdenum platform comprising a coordinated alkyne was employed for the cleavage of molecular dinitrogen. The coordinated η2‐alkyne was left unaffected during this reduction. ...DFT calculations suggest that the reaction proceeds via an initially generated terminal N2‐complex, which is converted to a dinuclear μ‐(η1:η1)‐N2‐bridged intermediate prior to N−N bond cleavage. Protonation, alkylation and acylation of the resulting molybdenum nitrido complex led to the corresponding N‐functionalized imido complexes. Upon oxidation of the N‐acylated imido derivative in MeCN, a fumaronitrile fragment was built up via C−C coupling of MeCN to afford a dinuclear molybdenum complex. The key finding that the strong N≡N bond may be cleaved in the presence of a weaker, but spatially constrained C≡C bond contradicts the widespread paradigm that coordinated alkynes are in general more reactive than gaseous N2.
A new molybdenum pincer for N2‐splitting and functionalization is reported. The N2‐cleavage affords a molybdenum nitrido complex and proceeds in the presence of a coordinated alkyne, that is, in the presence of a weaker C≡C bond. The spatially and rationally restricted alkyne moiety is also tolerated during N‐functionalization of the former nitrido complex.
Deep learning may transform health care, but model development has largely been dependent on availability of advanced technical expertise. Herein we present the development of a deep learning model ...by clinicians without coding, which predicts reported sex from retinal fundus photographs. A model was trained on 84,743 retinal fundus photos from the UK Biobank dataset. External validation was performed on 252 fundus photos from a tertiary ophthalmic referral center. For internal validation, the area under the receiver operating characteristic curve (AUROC) of the code free deep learning (CFDL) model was 0.93. Sensitivity, specificity, positive predictive value (PPV) and accuracy (ACC) were 88.8%, 83.6%, 87.3% and 86.5%, and for external validation were 83.9%, 72.2%, 78.2% and 78.6% respectively. Clinicians are currently unaware of distinct retinal feature variations between males and females, highlighting the importance of model explainability for this task. The model performed significantly worse when foveal pathology was present in the external validation dataset, ACC: 69.4%, compared to 85.4% in healthy eyes, suggesting the fovea is a salient region for model performance OR (95% CI): 0.36 (0.19, 0.70) p = 0.0022. Automated machine learning (AutoML) may enable clinician-driven automated discovery of novel insights and disease biomarkers.
PDF
