An algorithm to detect dicrotic notch in arterial blood pressure and photoplethysmography waveforms using the iterative envelope mean method.

BACKGROUND AND OBJECTIVE: Detection of the dicrotic notch (DN) within a cardiac cycle is essential for assessment of cardiac output, calculation of pulse wave velocity, estimation of left ventricular ejection time, and supporting feature-based machine learning models for noninvasive blood pressure estimation, and hypotension, or hypertension prediction. In this study, we present a new algorithm based on the iterative envelope mean (IEM) method to detect automatically the DN in arterial blood pressure (ABP) and photoplethysmography (PPG) waveforms. METHODS: The algorithm was evaluated on both ABP and PPG waveforms from a large perioperative dataset (MLORD dataset) comprising 17,327 patients. The analysis involved a total of 1,171,288 cardiac cycles for ABP waveforms and 3,424,975 cardiac cycles for PPG waveforms. To evaluate the algorithm's performance, the systolic phase duration (SPD) was employed, which represents the duration from the onset of the systolic phase to the DN in the cardiac cycle. Correlation plots and regression analysis were used to compare the algorithm against marked DN detection, while box plots and Bland-Altman plots were used to compare its performance with both marked DN detection and an established DN detection technique (second derivative). The marking of the DN temporal location was carried out by an experienced researcher using the help of the 'find_peaks' function from the scipy Python package, serving as a reference for the evaluation. The marking was visually validated by both an engineer and an anesthesiologist. The robustness of the algorithm was evaluated as the DN was made less visually distinct across signal-to-noise ratios (SNRs) ranging from -30 dB to -5 dB in both ABP and PPG waveforms. RESULTS: The correlation between SPD estimated by the algorithm and that marked by the researcher is strong for both ABP (R2(87,343) =0.99, p<.001) and PPG (R2(86,764) =0.98, p<.001) waveforms. The algorithm had a lower mean error of DN detection (s): 0.0047 (0.0029) for ABP waveforms and 0.0046 (0.0029) for PPG waveforms, compared to 0.0693 (0.0770) for ABP and 0.0968 (0.0909) for PPG waveforms for the established 2nd derivative method. The algorithm has high rate of detectability of DN detection for SNR of >= -9 dB for ABP waveforms and >= -12 dB for PPG waveforms indicating robust performance in detecting the DN when it is less visibly distinct. CONCLUSION: Our proposed IEM- based algorithm can detect DN in both ABP and PPG waveforms with low computational cost, even in cases where it is not distinctly defined within a cardiac cycle of the waveform ('DN-less signals'). The algorithm can potentially serve as a valuable, fast, and reliable tool for extracting features from ABP and PPG waveforms. It can be especially beneficial in medical applications where DN-based features, such as SPD, diastolic phase duration, and DN amplitude, play a significant role.

Predicting in-hospital mortality among patients admitted with a diagnosis of heart failure: a machine learning approach.

Existing risk prediction models for hospitalized heart failure patients are limited. We identified patients hospitalized with a diagnosis of heart failure between 7 May 2013 and 26 April 2022 from a large academic, quaternary care medical centre (training cohort). Demographics, medical comorbidities, vitals, and labs were collected and were used to construct random forest machine learning models to predict in-hospital mortality. Models were compared with logistic regression, and to commonly used heart failure risk scores. The models were subsequently validated in patients hospitalized with a diagnosis of heart failure from a second academic, community medical centre (validation cohort). The entire cohort comprised 21 802 patients, of which 14 539 were in the training cohort and 7263 were in the validation cohort. The median age (25th-75th percentile) was 70 (58-82) for the entire cohort, 43.2% were female, and 6.7% experienced inpatient mortality. In the overall cohort, 7621 (35.0%) patients had heart failure with reduced ejection fraction (EF ≤ 40%), 1271 (5.8%) had heart failure with mildly reduced EF (EF 41-49%), and 12 910 (59.2%) had heart failure with preserved EF (EF ≥ 50%). Random forest models in the validation cohort demonstrated a c-statistic (95% confidence interval) of 0.96 (0.95-0.97), sensitivity (SN) of 87.3%, and specificity (SP) of 90.6% for the prediction of in-hospital mortality. Models for those with HFrEF demonstrated a c-statistic of 0.96 (0.94-0.98), SN 88.2%, and SP 91.0%, and those for patients with HFpEF showed a c-statistic of 0.95 (0.93-0.97), SN 87.4%, and SP 89.5% for predicting in-hospital mortality. The random forest model significantly outperformed logistic regression (c-statistic 0.87, SN 75.9%, and SP 86.9%), and current existing risk scores including the Acute Decompensated Heart Failure National Registry risk score (c-statistic of 0.70, SN 69%, and SP 62%), and the Get With the Guidelines-Heart Failure risk score (c-statistic 0.69, SN 67%, and SP 63%); P < 0.001 for comparison. Machine learning models built from commonly recorded patient information can accurately predict in-hospital mortality among patients hospitalized with a diagnosis of heart failure.

An algorithm to detect dicrotic notch in arterial blood pressure and photoplethysmography waveforms using the iterative envelope mean method.

Background and Objective: Detection of the dicrotic notch (DN) within a cardiac cycle is essential for assessment of cardiac output, calculation of pulse wave velocity, estimation of left ventricular ejection time, and supporting feature-based machine learning models for noninvasive blood pressure estimation, and hypotension, or hypertension prediction. In this study, we present a new algorithm based on the iterative envelope mean (IEM) method to detect automatically the DN in arterial blood pressure (ABP) and photoplethysmography (PPG) waveforms. Methods: The algorithm was evaluated on both ABP and PPG waveforms from a large perioperative dataset (MLORD dataset) comprising 17,327 patients. The analysis involved a total of 1,171,288 cardiac cycles for ABP waveforms and 3,424,975 cardiac cycles for PPG waveforms. To evaluate the algorithm's performance, the systolic phase duration (SPD) was employed, which represents the duration from the onset of the systolic phase to the DN in the cardiac cycle. Correlation plots and regression analysis were used to compare the algorithm with an established DN detection technique (second derivative). The marking of the DN temporal location was carried out by an experienced researcher using the help of the 'find_peaks' function from the scipy PYTHON package, serving as a reference for the evaluation. The marking was visually validated by both an engineer and an anesthesiologist. The robustness of the algorithm was evaluated as the DN was made less visually distinct across signal-to-noise ratios (SNRs) ranging from -30 dB to -5 dB in both ABP and PPG waveforms. Results: The correlation between SPD estimated by the algorithm and that marked by the researcher is strong for both ABP (R2(87343) =.99, p<.001) and PPG (R2(86764) =.98, p<.001) waveforms. The algorithm had a lower mean error of dicrotic notch detection (s): 0.0047 (0.0029) for ABP waveforms and 0.0046 (0.0029) for PPG waveforms, compared to 0.0693 (0.0770) for ABP and 0.0968 (0.0909) for PPG waveforms for the established 2nd derivative method. The algorithm has high accuracy of DN detection for SNR of >= -9 dB for ABP waveforms and >= -12 dB for PPG waveforms indicating robust performance in detecting the DN when it is less visibly distinct. Conclusion: Our proposed IEM- based algorithm can detect DN in both ABP and PPG waveforms with low computational cost, even in cases where it is not distinctly defined within a cardiac cycle of the waveform ('DN-less signals'). The algorithm can potentially serve as a valuable, fast, and reliable tool for extracting features from ABP and PPG waveforms. It can be especially beneficial in medical applications where DN-based features, such as SPD, diastolic phase duration, and DN amplitude, play a significant role.

Antibiotic Switches in Urinary Tract Infection Are Associated With Atypical Symptoms and Emergent Care.

IMPORTANCE: Given worsening global antibiotic resistance, antimicrobial stewardship aims to use the shortest effective duration of the most narrow-spectrum, effective antibiotic for patients with specific urinary symptoms and laboratory testing consistent with urinary tract infection (UTI). Inappropriate treatment and unnecessary antibiotic switching for UTIs harms patients in a multitude of ways. OBJECTIVE: This study sought to analyze antibiotic treatment failures as measured by antibiotic switching for treatment of UTI in emergent and ambulatory care. STUDY DESIGN: For this retrospective cohort study, 908 encounters during July 2019 bearing a diagnostic code for UTI/cystitis in a single health care system were reviewed. Urinary and microbiological testing, symptoms endorsed at presentation, and treatments prescribed were extracted from the medical record. RESULTS: Of 908 patients diagnosed with UTI, 64% of patients (579/908) received antibiotics, 86% of which were empiric. All patients evaluated in emergent care settings were prescribed antibiotics empirically in contrast to 71% of patients in ambulatory settings (P < 0.001). Of patients given antibiotics, 89 of 579 patients (15%, 10% of all 908 patients) were switched to alternative antibiotics within 28 days. Emergent care settings and positive urine cultures were significantly associated with increased antibiotic switching. Patients subjected to switching tended to have higher rates of presenting symptoms inconsistent with UTI. CONCLUSIONS: Empiric treatment, particularly in an emergent care setting, was frequently inappropriate and associated with increasing rates of antibiotic switching. Given the profound potential contribution to antibiotic resistance, these findings highlight the need for improved diagnostic and prescribing accuracy for UTI.

SLIViT: a general AI framework for clinical-feature diagnosis from limited 3D biomedical-imaging data.

We present SLIViT, a deep-learning framework that accurately measures disease-related risk factors in volumetric biomedical imaging, such as magnetic resonance imaging (MRI) scans, optical coherence tomography (OCT) scans, and ultrasound videos. To evaluate SLIViT, we applied it to five different datasets of these three different data modalities tackling seven learning tasks (including both classification and regression) and found that it consistently and significantly outperforms domain-specific state-of-the-art models, typically improving performance (ROC AUC or correlation) by 0.1-0.4. Notably, compared to existing approaches, SLIViT can be applied even when only a small number of annotated training samples is available, which is often a constraint in medical applications. When trained on less than 700 annotated volumes, SLIViT obtained accuracy comparable to trained clinical specialists while reducing annotation time by a factor of 5,000 demonstrating its utility to automate and expedite ongoing research and other practical clinical scenarios.

Ice Cores as a Source for Antimicrobials: From Bioprospecting to Biodesign.

The golden age has passed for antibiotic discovery, and while some antibiotics are currently in various phases of clinical trials in the United States, many pharmaceutical companies have abandoned antibiotic research. With the need for antibiotics, we should expand our horizon for therapeutic mining and can look toward understudied sources such as ice cores. Ice cores contain microorganisms and genetic material that have been frozen in time for thousands of years. The antibiotics used by these organisms are encoded in their genomes, which can be unlocked, identified, and characterized with modern advances in molecular biology, genetic sequencing, various computational approaches, and established natural product discovery pipelines. While synthetic biology can be used in natural product discovery approaches, synthetic biology and bioengineering efforts can also be leveraged in the selection and biodesign of increased compound yields, potency, and stability. Here, we provide the perspective that ice cores can be a source of novel antibiotic compounds and that the tools of synthetic biology can be used to design better antimicrobials.

TRAF6 and TRAF2/3 Binding Motifs in CD40 Differentially Regulate B Cell Function in T-Dependent Antibody Responses and Dendritic Cell Function in Experimental Autoimmune Encephalomyelitis.

Expression of the costimulatory molecule CD40 on both B cells and dendritic cells (DCs) is required for induction of experimental autoimmune encephalomyelitis (EAE), and cell-autonomous CD40 expression on B cells is required for primary T-dependent (TD) Ab responses. We now ask whether the function of CD40 expressed by different cell types in these responses is mediated by the same or different cytoplasmic domains. CD40 has been reported to possess multiple cytoplasmic domains, including distinct TRAF6 and TRAF2/3 binding motifs. To elucidate the in vivo function of these motifs in B cells and DCs involved in EAE and TD germinal center responses, we have generated knock-in mice containing distinct CD40 cytoplasmic domain TRAF-binding site mutations and have used these animals, together with bone marrow chimeric mice, to assess the roles that these motifs play in CD40 function. We found that both TRAF2/3 and TRAF6 motifs of CD40 are critically involved in EAE induction and demonstrated that this is mediated by a role of both motifs for priming of pathogenic T cells by DCs. In contrast, the TRAF2/3 binding motif, but not the TRAF6 binding motif, is required for B cell CD40 function in TD high-affinity Ab responses. These data demonstrate that the requirements for expression of specific TRAF-binding CD40 motifs differ for B cells or DCs that function in specific immune responses and thus identify targets for intervention to modulate these responses.

Re-Imagining Public Health Education: Students' Perspectives and Considerations.

Since the onset of the COVID-19 pandemic, the shortcomings and neglected weaknesses of public health systems have risen to the surface, emphasizing the need for new approaches to designing and delivering public health training. Higher education institutions have a critical role in advocating for societal change and sufficiently prepare the next generations of public health. Therefore, this commentary shares the unique voices of current and recently completed graduate students from public health programs across the United States in identifying areas of improvement, so that proactive steps toward refining the current landscape of public health education and training may be taken. We speak upon the inaction and accountability of public health academic spaces in dismantling the various forms of systemic oppression, such as racism, colonialism, and epistemic injustice, while encouraging prospective and current graduate colleagues to be mindful and curious to re-imagine the role of such pedagogies of public health to reduce the progressing health inequities.

Association of red blood cell distribution width with hospital admission and in-hospital mortality across all-cause adult emergency department visits.

Objectives: To test the association between the initial red blood cell distribution width (RDW) value in the emergency department (ED) and hospital admission and, among those admitted, in-hospital mortality. Materials and Methods: We perform a retrospective analysis of 210 930 adult ED visits with complete blood count results from March 2013 to February 2022. Primary outcomes were hospital admission and in-hospital mortality. Variables for each visit included demographics, comorbidities, vital signs, basic metabolic panel, complete blood count, and final diagnosis. The association of each outcome with the initial RDW value was calculated across 3 age groups (<45, 45-65, and >65) as well as across 374 diagnosis categories. Logistic regression (LR) and XGBoost models using all variables excluding final diagnoses were built to test whether RDW was a highly weighted and informative predictor for each outcome. Finally, simplified models using only age, sex, and vital signs were built to test whether RDW had additive predictive value. Results: Compared to that of discharged visits (mean [SD]: 13.8 [2.03]), RDW was significantly elevated in visits that resulted in admission (15.1 [2.72]) and, among admissions, those resulting in intensive care unit stay (15.3 [2.88]) and/or death (16.8 [3.25]). This relationship held across age groups as well as across various diagnosis categories. An RDW >16 achieved 90% specificity for hospital admission, while an RDW >18.5 achieved 90% specificity for in-hospital mortality. LR achieved a test area under the curve (AUC) of 0.77 (95% confidence interval [CI] 0.77-0.78) for hospital admission and 0.85 (95% CI 0.81-0.88) for in-hospital mortality, while XGBoost achieved a test AUC of 0.90 (95% CI 0.89-0.90) for hospital admission and 0.96 (95% CI 0.94-0.97) for in-hospital mortality. RDW had high scaled weights and information gain for both outcomes and had additive value in simplified models predicting hospital admission. Discussion: Elevated RDW, previously associated with mortality in myocardial infarction, pulmonary embolism, heart failure, sepsis, and COVID-19, is associated with hospital admission and in-hospital mortality across all-cause adult ED visits. Used alone, elevated RDW may be a specific, but not sensitive, test for both outcomes, with multivariate LR and XGBoost models showing significantly improved test characteristics. Conclusions: RDW, a component of the complete blood count panel routinely ordered as the initial workup for the undifferentiated patient, may be a generalizable biomarker for acuity in the ED.

Effect of statins on the age of onset of age-related macular degeneration.

BACKGROUND: This study evaluated the relationship between statin use and the age of onset of age-related macular degeneration (AMD). METHODS: Electronic Health Records from 52,840 patients evaluated at University of California Los Angeles (UCLA) Ophthalmology Clinics and 9,977 patients evaluated at University of California San Francisco (UCSF) Ophthalmology Clinics were screened. Survival analysis was performed using Cox proportional hazards regression models and visualized using Kaplan Meier survival curves, with the following covariates-sex, ethnicity, smoking history, fluoxetine use, obesity, diabetes mellitus, and hypertension. RESULTS: 5,498 of 52,840 patients at UCLA were diagnosed with AMD. Statin use was associated with a later AMD onset (HR = 0.8823, p < 0.0001), while female sex (HR = 1.0852, p= 00,035), obesity (HR = 1.4555, p < 0.0001), and fluoxetine (HR = 1.3797, p= 0.0003) were associated with an earlier AMD onset. Non-hispanic black (HR = 0.5687, p < 0.0001) and hispanic ethnicities (HR = 0.8269, p= 0.0028) were associated with a later AMD onset. When stratifying for ethnicity, statins, fluoxetine, sex, and obesity were significant only within non-hispanic white subjects. Statin use was significant among patients with dry AMD (HR = 0.8410, p= 0.0001) but not wet AMD (0.9188, p= 0.0351). In the replication cohort, 526 of 9,977 patients at UCSF had AMD. Associations between statins (HR = 0.7643, p= 0.0033), non-hispanic black ethnicity (HR = 0.5043, p= 0.0035), and obesity (HR = 1.9602, p < 0.0001) on AMD onset were confirmed. CONCLUSIONS: In both cohorts, statin use and non-hispanic black ethnicity are associated with a later AMD onset, while obesity with an earlier AMD onset.

Automated large-scale prediction of exudative AMD progression using machine-read OCT biomarkers.

Age-related Macular Degeneration (AMD) is a major cause of irreversible vision loss in individuals over 55 years old in the United States. One of the late-stage manifestations of AMD, and a major cause of vision loss, is the development of exudative macular neovascularization (MNV). Optical Coherence Tomography (OCT) is the gold standard to identify fluid at different levels within the retina. The presence of fluid is considered the hallmark to define the presence of disease activity. Anti-vascular growth factor (anti-VEGF) injections can be used to treat exudative MNV. However, given the limitations of anti-VEGF treatment, as burdensome need for frequent visits and repeated injections to sustain efficacy, limited durability of the treatment, poor or no response, there is a great interest in detecting early biomarkers associated with a higher risk for AMD progression to exudative forms in order to optimize the design of early intervention clinical trials. The annotation of structural biomarkers on optical coherence tomography (OCT) B-scans is a laborious, complex and time-consuming process, and discrepancies between human graders can introduce variability into this assessment. To address this issue, a deep-learning model (SLIVER-net) was proposed, which could identify AMD biomarkers on structural OCT volumes with high precision and without human supervision. However, the validation was performed on a small dataset, and the true predictive power of these detected biomarkers in the context of a large cohort has not been evaluated. In this retrospective cohort study, we perform the largest-scale validation of these biomarkers to date. We also assess how these features combined with other EHR data (demographics, comorbidities, etc) affect and/or improve the prediction performance relative to known factors. Our hypothesis is that these biomarkers can be identified by a machine learning algorithm without human supervision, in a way that they preserve their predictive nature. The way we test this hypothesis is by building several machine learning models utilizing these machine-read biomarkers and assessing their added predictive power. We found that not only can we show that the machine-read OCT B-scan biomarkers are predictive of AMD progression, we also observe that our proposed combined OCT and EHR data-based algorithm outperforms the state-of-the-art solution in clinically relevant metrics and provides actionable information which has the potential to improve patient care. In addition, it provides a framework for automated large-scale processing of OCT volumes, making it possible to analyze vast archives without human supervision.

Automated Identification of Incomplete and Complete Retinal Epithelial Pigment and Outer Retinal Atrophy Using Machine Learning.

OBJECTIVE: To assess and validate a deep learning algorithm to automatically detect incomplete retinal pigment epithelial and outer retinal atrophy (iRORA) and complete retinal pigment epithelial and outer retinal atrophy (cRORA) in eyes with age-related macular degeneration. DESIGN: In a retrospective machine learning analysis, a deep learning model was trained to jointly classify the presence of iRORA and cRORA within a given B-scan. The algorithm was evaluated using 2 separate and independent datasets. PARTICIPANTS: OCT B-scan volumes from 71 patients with nonneovascular age-related macular degeneration captured at the Doheny-University of California Los Angeles Eye Centers and the following 2 external OCT B-scans testing datasets: (1) University of Pennsylvania, University of Miami, and Case Western Reserve University and (2) Doheny Image Reading Research Laboratory. METHODS: The images were annotated by an experienced grader for the presence of iRORA and cRORA. A Resnet18 model was trained to classify these annotations for each B-scan using OCT volumes collected at the Doheny-University of California Los Angeles Eye Centers. The model was applied to 2 testing datasets to assess out-of-sample model performance. MAIN OUTCOMES MEASURES: Model performance was quantified in terms of area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC). Sensitivity, specificity, and positive predictive value were also compared against additional clinician annotators. RESULTS: On an independently collected test set, consisting of 1117 volumes from the general population, the model predicted iRORA and cRORA presence within the entire volume with nearly perfect AUROC performance and AUPRC scores (iRORA, 0.61; 95% confidence interval [CI] [0.45, 0.82]: cRORA, 0.83; 95% CI [0.68, 0.95]). On another independently collected set, consisting of 60 OCT B-scans enriched for iRORA and cRORA lesions, the model performed with AUROC (iRORA: 0.68, 95% CI [0.54, 0.81]; cRORA: 0.84, 95% CI [0.75, 0.94]) and AUPRC (iRORA: 0.70, 95% CI [0.55, 0.86]; cRORA: 0.82, 95% CI [0.70, 0.93]). CONCLUSIONS: A deep learning model can accurately and precisely identify both iRORA and cRORA lesions within the OCT B-scan volume. The model can achieve similar sensitivity compared with human graders, which potentially obviates a laborious and time-consuming annotation process and could be developed into a diagnostic screening tool.

Neural complexity is a common denominator of human consciousness across diverse regimes of cortical dynamics.

What is the common denominator of consciousness across divergent regimes of cortical dynamics? Does consciousness show itself in decibels or in bits? To address these questions, we introduce a testbed for evaluating electroencephalogram (EEG) biomarkers of consciousness using dissociations between neural oscillations and consciousness caused by rare genetic disorders. Children with Angelman syndrome (AS) exhibit sleep-like neural dynamics during wakefulness. Conversely, children with duplication 15q11.2-13.1 syndrome (Dup15q) exhibit wake-like neural dynamics during non-rapid eye movement (NREM) sleep. To identify highly generalizable biomarkers of consciousness, we trained regularized logistic regression classifiers on EEG data from wakefulness and NREM sleep in children with AS using both entropy measures of neural complexity and spectral (i.e., neural oscillatory) EEG features. For each set of features, we then validated these classifiers using EEG from neurotypical (NT) children and abnormal EEGs from children with Dup15q. Our results show that the classification performance of entropy-based EEG biomarkers of conscious state is not upper-bounded by that of spectral EEG features, which are outperformed by entropy features. Entropy-based biomarkers of consciousness may thus be highly adaptable and should be investigated further in situations where spectral EEG features have shown limited success, such as detecting covert consciousness or anesthesia awareness.

Extended reality as a means to enhance public health education.

Technology has transformed the classroom and learning environments. From electronic whiteboards to tablet computers, educators now have access to a multitude of tools that enhance the learning experience. Educational technologies that rely on extended reality (XR) such as augmented and virtual reality are being used, or suggested for use, in various settings and often focus on technical fields such as medicine, dentistry, and aviation. Here, we propose that XR can be used in public health education to better prepare both undergraduate and graduate trainees for real world, complex public health scenarios that require public engagement, investigative skills, and critical decision making. Several opportunities for XR use are outlined that provide perspective on how XR can supplement traditional classroom instruction methods by providing an immersive, participatory training environment. XR offers an opportunity for public health students to gain confidence, have repeated simulated exposures in a safe and equitable environment, and build competency in critical functions they will likely perform as future public health professionals.

Methylation risk scores are associated with a collection of phenotypes within electronic health record systems.

Inference of clinical phenotypes is a fundamental task in precision medicine, and has therefore been heavily investigated in recent years in the context of electronic health records (EHR) using a large arsenal of machine learning techniques, as well as in the context of genetics using polygenic risk scores (PRS). In this work, we considered the epigenetic analog of PRS, methylation risk scores (MRS), a linear combination of methylation states. We measured methylation across a large cohort (n = 831) of diverse samples in the UCLA Health biobank, for which both genetic and complete EHR data are available. We constructed MRS for 607 phenotypes spanning diagnoses, clinical lab tests, and medication prescriptions. When added to a baseline set of predictive features, MRS significantly improved the imputation of 139 outcomes, whereas the PRS improved only 22 (median improvement for methylation 10.74%, 141.52%, and 15.46% in medications, labs, and diagnosis codes, respectively, whereas genotypes only improved the labs at a median increase of 18.42%). We added significant MRS to state-of-the-art EHR imputation methods that leverage the entire set of medical records, and found that including MRS as a medical feature in the algorithm significantly improves EHR imputation in 37% of lab tests examined (median R2 increase 47.6%). Finally, we replicated several MRS in multiple external studies of methylation (minimum p-value of 2.72 × 10-7) and replicated 22 of 30 tested MRS internally in two separate cohorts of different ethnicity. Our publicly available results and weights show promise for methylation risk scores as clinical and scientific tools.

Imputation of the continuous arterial line blood pressure waveform from non-invasive measurements using deep learning.

In two-thirds of intensive care unit (ICU) patients and 90% of surgical patients, arterial blood pressure (ABP) is monitored non-invasively but intermittently using a blood pressure cuff. Since even a few minutes of hypotension increases the risk of mortality and morbidity, for the remaining (high-risk) patients ABP is measured continuously using invasive devices, and derived values are extracted from the recorded waveforms. However, since invasive monitoring is associated with major complications (infection, bleeding, thrombosis), the ideal ABP monitor should be both non-invasive and continuous. With large volumes of high-fidelity physiological waveforms, it may be possible today to impute a physiological waveform from other available signals. Currently, the state-of-the-art approaches for ABP imputation only aim at intermittent systolic and diastolic blood pressure imputation, and there is no method that imputes the continuous ABP waveform. Here, we developed a novel approach to impute the continuous ABP waveform non-invasively using two continuously-monitored waveforms that are currently part of the standard-of-care, the electrocardiogram (ECG) and photo-plethysmogram (PPG), by adapting a deep learning architecture designed for image segmentation. Using over 150,000 min of data collected at two separate health systems from 463 patients, we demonstrate that our model provides a highly accurate prediction of the continuous ABP waveform (root mean square error 5.823 (95% CI 5.806-5.840) mmHg), as well as the derived systolic (mean difference 2.398 ± 5.623 mmHg) and diastolic blood pressure (mean difference - 2.497 ± 3.785 mmHg) compared to arterial line measurements. Our approach can potentially be used to measure blood pressure continuously and non-invasively for all patients in the acute care setting, without the need for any additional instrumentation beyond the current standard-of-care.

Automated identification of clinical features from sparsely annotated 3-dimensional medical imaging.

One of the core challenges in applying machine learning and artificial intelligence to medicine is the limited availability of annotated medical data. Unlike in other applications of machine learning, where an abundance of labeled data is available, the labeling and annotation of medical data and images require a major effort of manual work by expert clinicians who do not have the time to annotate manually. In this work, we propose a new deep learning technique (SLIVER-net), to predict clinical features from 3-dimensional volumes using a limited number of manually annotated examples. SLIVER-net is based on transfer learning, where we borrow information about the structure and parameters of the network from publicly available large datasets. Since public volume data are scarce, we use 2D images and account for the 3-dimensional structure using a novel deep learning method which tiles the volume scans, and then adds layers that leverage the 3D structure. In order to illustrate its utility, we apply SLIVER-net to predict risk factors for progression of age-related macular degeneration (AMD), a leading cause of blindness, from optical coherence tomography (OCT) volumes acquired from multiple sites. SLIVER-net successfully predicts these factors despite being trained with a relatively small number of annotated volumes (hundreds) and only dozens of positive training examples. Our empirical evaluation demonstrates that SLIVER-net significantly outperforms standard state-of-the-art deep learning techniques used for medical volumes, and its performance is generalizable as it was validated on an external testing set. In a direct comparison with a clinician panel, we find that SLIVER-net also outperforms junior specialists, and identifies AMD progression risk factors similarly to expert retina specialists.