Blinded, randomized trial of sonographer versus AI cardiac function assessment.

Artificial intelligence (AI) has been developed for echocardiography1-3, although it has not yet been tested with blinding and randomization. Here we designed a blinded, randomized non-inferiority clinical trial (ClinicalTrials.gov ID: NCT05140642; no outside funding) of AI versus sonographer initial assessment of left ventricular ejection fraction (LVEF) to evaluate the impact of AI in the interpretation workflow. The primary end point was the change in the LVEF between initial AI or sonographer assessment and final cardiologist assessment, evaluated by the proportion of studies with substantial change (more than 5% change). From 3,769 echocardiographic studies screened, 274 studies were excluded owing to poor image quality. The proportion of studies substantially changed was 16.8% in the AI group and 27.2% in the sonographer group (difference of -10.4%, 95% confidence interval: -13.2% to -7.7%, P < 0.001 for non-inferiority, P < 0.001 for superiority). The mean absolute difference between final cardiologist assessment and independent previous cardiologist assessment was 6.29% in the AI group and 7.23% in the sonographer group (difference of -0.96%, 95% confidence interval: -1.34% to -0.54%, P < 0.001 for superiority). The AI-guided workflow saved time for both sonographers and cardiologists, and cardiologists were not able to distinguish between the initial assessments by AI versus the sonographer (blinding index of 0.088). For patients undergoing echocardiographic quantification of cardiac function, initial assessment of LVEF by AI was non-inferior to assessment by sonographers.

Genome-wide chromatin state transitions associated with developmental and environmental cues.

Differences in chromatin organization are key to the multiplicity of cell states that arise from a single genetic background, yet the landscapes of in vivo tissues remain largely uncharted. Here, we mapped chromatin genome-wide in a large and diverse collection of human tissues and stem cells. The maps yield unprecedented annotations of functional genomic elements and their regulation across developmental stages, lineages, and cellular environments. They also reveal global features of the epigenome, related to nuclear architecture, that also vary across cellular phenotypes. Specifically, developmental specification is accompanied by progressive chromatin restriction as the default state transitions from dynamic remodeling to generalized compaction. Exposure to serum in vitro triggers a distinct transition that involves de novo establishment of domains with features of constitutive heterochromatin. We describe how these global chromatin state transitions relate to chromosome and nuclear architecture, and discuss their implications for lineage fidelity, cellular senescence, and reprogramming.

Video-based AI for beat-to-beat assessment of cardiac function.

Accurate assessment of cardiac function is crucial for the diagnosis of cardiovascular disease1, screening for cardiotoxicity2 and decisions regarding the clinical management of patients with a critical illness3. However, human assessment of cardiac function focuses on a limited sampling of cardiac cycles and has considerable inter-observer variability despite years of training4,5. Here, to overcome this challenge, we present a video-based deep learning algorithm-EchoNet-Dynamic-that surpasses the performance of human experts in the critical tasks of segmenting the left ventricle, estimating ejection fraction and assessing cardiomyopathy. Trained on echocardiogram videos, our model accurately segments the left ventricle with a Dice similarity coefficient of 0.92, predicts ejection fraction with a mean absolute error of 4.1% and reliably classifies heart failure with reduced ejection fraction (area under the curve of 0.97). In an external dataset from another healthcare system, EchoNet-Dynamic predicts the ejection fraction with a mean absolute error of 6.0% and classifies heart failure with reduced ejection fraction with an area under the curve of 0.96. Prospective evaluation with repeated human measurements confirms that the model has variance that is comparable to or less than that of human experts. By leveraging information across multiple cardiac cycles, our model can rapidly identify subtle changes in ejection fraction, is more reproducible than human evaluation and lays the foundation for precise diagnosis of cardiovascular disease in real time. As a resource to promote further innovation, we also make publicly available a large dataset of 10,030 annotated echocardiogram videos.

Classification and clustering of RNA crosslink-ligation data reveal complex structures and homodimers.

The recent development and application of methods based on the general principle of "crosslinking and proximity ligation" (crosslink-ligation) are revolutionizing RNA structure studies in living cells. However, extracting structure information from such data presents unique challenges. Here, we introduce a set of computational tools for the systematic analysis of data from a wide variety of crosslink-ligation methods, specifically focusing on read mapping, alignment classification, and clustering. We design a new strategy to map short reads with irregular gaps at high sensitivity and specificity. Analysis of previously published data reveals distinct properties and bias caused by the crosslinking reactions. We perform rigorous and exhaustive classification of alignments and discover eight types of arrangements that provide distinct information on RNA structures and interactions. To deconvolve the dense and intertwined gapped alignments, we develop a network/graph-based tool Crosslinked RNA Secondary Structure Analysis using Network Techniques (CRSSANT), which enables clustering of gapped alignments and discovery of new alternative and dynamic conformations. We discover that multiple crosslinking and ligation events can occur on the same RNA, generating multisegment alignments to report complex high-level RNA structures and multi-RNA interactions. We find that alignments with overlapped segments are produced from potential homodimers and develop a new method for their de novo identification. Analysis of overlapping alignments revealed potential new homodimers in cellular noncoding RNAs and RNA virus genomes in the Picornaviridae family. Together, this suite of computational tools enables rapid and efficient analysis of RNA structure and interaction data in living cells.

Machine learning modeling of RNA structures: methods, challenges and future perspectives.

The three-dimensional structure of RNA molecules plays a critical role in a wide range of cellular processes encompassing functions from riboswitches to epigenetic regulation. These RNA structures are incredibly dynamic and can indeed be described aptly as an ensemble of structures that shifts in distribution depending on different cellular conditions. Thus, the computational prediction of RNA structure poses a unique challenge, even as computational protein folding has seen great advances. In this review, we focus on a variety of machine learning-based methods that have been developed to predict RNA molecules' secondary structure, as well as more complex tertiary structures. We survey commonly used modeling strategies, and how many are inspired by or incorporate thermodynamic principles. We discuss the shortcomings that various design decisions entail and propose future directions that could build off these methods to yield more robust, accurate RNA structure predictions.

Clinical Genetics Lacks Standard Definitions and Protocols for the Collection and Use of Diversity Measures.

Genetics researchers and clinical professionals rely on diversity measures such as race, ethnicity, and ancestry (REA) to stratify study participants and patients for a variety of applications in research and precision medicine. However, there are no comprehensive, widely accepted standards or guidelines for collecting and using such data in clinical genetics practice. Two NIH-funded research consortia, the Clinical Genome Resource (ClinGen) and Clinical Sequencing Evidence-generating Research (CSER), have partnered to address this issue and report how REA are currently collected, conceptualized, and used. Surveying clinical genetics professionals and researchers (n = 448), we found heterogeneity in the way REA are perceived, defined, and measured, with variation in the perceived importance of REA in both clinical and research settings. The majority of respondents (>55%) felt that REA are at least somewhat important for clinical variant interpretation, ordering genetic tests, and communicating results to patients. However, there was no consensus on the relevance of REA, including how each of these measures should be used in different scenarios and what information they can convey in the context of human genetics. A lack of common definitions and applications of REA across the precision medicine pipeline may contribute to inconsistencies in data collection, missing or inaccurate classifications, and misleading or inconclusive results. Thus, our findings support the need for standardization and harmonization of REA data collection and use in clinical genetics and precision health research.

Deep profiling of protease substrate specificity enabled by dual random and scanned human proteome substrate phage libraries.

Proteolysis is a major posttranslational regulator of biology inside and outside of cells. Broad identification of optimal cleavage sites and natural substrates of proteases is critical for drug discovery and to understand protease biology. Here, we present a method that employs two genetically encoded substrate phage display libraries coupled with next generation sequencing (SPD-NGS) that allows up to 10,000-fold deeper sequence coverage of the typical six- to eight-residue protease cleavage sites compared to state-of-the-art synthetic peptide libraries or proteomics. We applied SPD-NGS to two classes of proteases, the intracellular caspases, and the ectodomains of the sheddases, ADAMs 10 and 17. The first library (Lib 10AA) allowed us to identify 104 to 105 unique cleavage sites over a 1,000-fold dynamic range of NGS counts and produced consensus and optimal cleavage motifs based position-specific scoring matrices. A second SPD-NGS library (Lib hP), which displayed virtually the entire human proteome tiled in contiguous 49 amino acid sequences with 25 amino acid overlaps, enabled us to identify candidate human proteome sequences. We identified up to 104 natural linear cut sites, depending on the protease, and captured most of the examples previously identified by proteomics and predicted 10- to 100-fold more. Structural bioinformatics was used to facilitate the identification of candidate natural protein substrates. SPD-NGS is rapid, reproducible, simple to perform and analyze, inexpensive, and renewable, with unprecedented depth of coverage for substrate sequences, and is an important tool for protease biologists interested in protease specificity for specific assays and inhibitors and to facilitate identification of natural protein substrates.

Genetic and socioeconomic study of mate choice in Latinos reveals novel assortment patterns.

Nonrandom mating in human populations has important implications for genetics and medicine as well as for economics and sociology. In this study, we performed an integrative analysis of a large cohort of Mexican and Puerto Rican couples using detailed socioeconomic attributes and genotypes. We found that in ethnically homogeneous Latino communities, partners are significantly more similar in their genomic ancestries than expected by chance. Consistent with this, we also found that partners are more closely related--equivalent to between third and fourth cousins in Mexicans and Puerto Ricans--than matched random male-female pairs. Our analysis showed that this genomic ancestry similarity cannot be explained by the standard socioeconomic measurables alone. Strikingly, the assortment of genomic ancestry in couples was consistently stronger than even the assortment of education. We found enriched correlation of partners' genotypes at genes known to be involved in facial development. We replicated our results across multiple geographic locations. We discuss the implications of assortment and assortment-specific loci on disease dynamics and disease mapping methods in Latinos.

Inferring parental genomic ancestries using pooled semi-Markov processes.

MOTIVATION: A basic problem of broad public and scientific interest is to use the DNA of an individual to infer the genomic ancestries of the parents. In particular, we are often interested in the fraction of each parent's genome that comes from specific ancestries (e.g. European, African, Native American, etc). This has many applications ranging from understanding the inheritance of ancestry-related risks and traits to quantifying human assortative mating patterns. RESULTS: We model the problem of parental genomic ancestry inference as a pooled semi-Markov process. We develop a general mathematical framework for pooled semi-Markov processes and construct efficient inference algorithms for these models. Applying our inference algorithm to genotype data from 231 Mexican trios and 258 Puerto Rican trios where we have the true genomic ancestry of each parent, we demonstrate that our method accurately infers parameters of the semi-Markov processes and parents' genomic ancestries. We additionally validated the method on simulations. Our model of pooled semi-Markov process and inference algorithms may be of independent interest in other settings in genomics and machine learning.

Protein structure generation via folding diffusion.

The ability to computationally generate novel yet physically foldable protein structures could lead to new biological discoveries and new treatments targeting yet incurable diseases. Despite recent advances in protein structure prediction, directly generating diverse, novel protein structures from neural networks remains difficult. In this work, we present a diffusion-based generative model that generates protein backbone structures via a procedure inspired by the natural folding process. We describe a protein backbone structure as a sequence of angles capturing the relative orientation of the constituent backbone atoms, and generate structures by denoising from a random, unfolded state towards a stable folded structure. Not only does this mirror how proteins natively twist into energetically favorable conformations, the inherent shift and rotational invariance of this representation crucially alleviates the need for more complex equivariant networks. We train a denoising diffusion probabilistic model with a simple transformer backbone and demonstrate that our resulting model unconditionally generates highly realistic protein structures with complexity and structural patterns akin to those of naturally-occurring proteins. As a useful resource, we release an open-source codebase and trained models for protein structure diffusion.

Electrocardiographic deep learning for predicting post-procedural mortality: a model development and validation study.

BACKGROUND: Preoperative risk assessments used in clinical practice are insufficient in their ability to identify risk for postoperative mortality. Deep-learning analysis of electrocardiography can identify hidden risk markers that can help to prognosticate postoperative mortality. We aimed to develop a prognostic model that accurately predicts postoperative mortality in patients undergoing medical procedures and who had received preoperative electrocardiographic diagnostic testing. METHODS: In a derivation cohort of preoperative patients with available electrocardiograms (ECGs) from Cedars-Sinai Medical Center (Los Angeles, CA, USA) between Jan 1, 2015 and Dec 31, 2019, a deep-learning algorithm was developed to leverage waveform signals to discriminate postoperative mortality. We randomly split patients (8:1:1) into subsets for training, internal validation, and final algorithm test analyses. Model performance was assessed using area under the receiver operating characteristic curve (AUC) values in the hold-out test dataset and in two external hospital cohorts and compared with the established Revised Cardiac Risk Index (RCRI) score. The primary outcome was post-procedural mortality across three health-care systems. FINDINGS: 45 969 patients had a complete ECG waveform image available for at least one 12-lead ECG performed within the 30 days before the procedure date (59 975 inpatient procedures and 112 794 ECGs): 36 839 patients in the training dataset, 4549 in the internal validation dataset, and 4581 in the internal test dataset. In the held-out internal test cohort, the algorithm discriminates mortality with an AUC value of 0·83 (95% CI 0·79-0·87), surpassing the discrimination of the RCRI score with an AUC of 0·67 (0·61-0·72). The algorithm similarly discriminated risk for mortality in two independent US health-care systems, with AUCs of 0·79 (0·75-0·83) and 0·75 (0·74-0·76), respectively. Patients determined to be high risk by the deep-learning model had an unadjusted odds ratio (OR) of 8·83 (5·57-13·20) for postoperative mortality compared with an unadjusted OR of 2·08 (0·77-3·50) for postoperative mortality for RCRI scores of more than 2. The deep-learning algorithm performed similarly for patients undergoing cardiac surgery (AUC 0·85 [0·77-0·92]), non-cardiac surgery (AUC 0·83 [0·79-0·88]), and catheterisation or endoscopy suite procedures (AUC 0·76 [0·72-0·81]). INTERPRETATION: A deep-learning algorithm interpreting preoperative ECGs can improve discrimination of postoperative mortality. The deep-learning algorithm worked equally well for risk stratification of cardiac surgeries, non-cardiac surgeries, and catheterisation laboratory procedures, and was validated in three independent health-care systems. This algorithm can provide additional information to clinicians making the decision to perform medical procedures and stratify the risk of future complications. FUNDING: National Heart, Lung, and Blood Institute.

Video-Based Deep Learning for Automated Assessment of Left Ventricular Ejection Fraction in Pediatric Patients.

BACKGROUND: Significant interobserver and interstudy variability occurs for left ventricular (LV) functional indices despite standardization of measurement techniques. Artificial intelligence models trained on adult echocardiograms are not likely to be applicable to a pediatric population. We present EchoNet-Peds, a video-based deep learning algorithm, which matches human expert performance of LV segmentation and ejection fraction (EF). METHODS: A large pediatric data set of 4,467 echocardiograms was used to develop EchoNet-Peds. EchoNet-Peds was trained on 80% of the data for segmentation of the left ventricle and estimation of LVEF. The remaining 20% was used to fine-tune and validate the algorithm. RESULTS: In both apical 4-chamber and parasternal short-axis views, EchoNet-Peds segments the left ventricle with a Dice similarity coefficient of 0.89. EchoNet-Peds estimates EF with a mean absolute error of 3.66% and can routinely identify pediatric patients with systolic dysfunction (area under the curve of 0.95). EchoNet-Peds was trained on pediatric echocardiograms and performed significantly better to estimate EF (P < .001) than an adult model applied to the same data. CONCLUSIONS: Accurate, rapid automation of EF assessment and recognition of systolic dysfunction in a pediatric population are feasible using EchoNet-Peds with the potential for far-reaching clinical impact. In addition, the first large pediatric data set of annotated echocardiograms is now publicly available for efforts to develop pediatric-specific artificial intelligence algorithms.

Understanding the molecular basis of resilience to Alzheimer's disease.

The cellular and molecular distinction between brain aging and neurodegenerative disease begins to blur in the oldest old. Approximately 15-25% of observations in humans do not fit predicted clinical manifestations, likely the result of suppressed damage despite usually adequate stressors and of resilience, the suppression of neurological dysfunction despite usually adequate degeneration. Factors during life may predict the clinico-pathologic state of resilience: cardiovascular health and mental health, more so than educational attainment, are predictive of a continuous measure of resilience to Alzheimer's disease (AD) and AD-related dementias (ADRDs). In resilience to AD alone (RAD), core features include synaptic and axonal processes, especially in the hippocampus. Future focus on larger and more diverse cohorts and additional regions offer emerging opportunities to understand this counterforce to neurodegeneration. The focus of this review is the molecular basis of resilience to AD.

Brain proteomic analysis implicates actin filament processes and injury response in resilience to Alzheimer's disease.

Resilience to Alzheimer's disease is an uncommon combination of high disease burden without dementia that offers valuable insights into limiting clinical impact. Here we assessed 43 research participants meeting stringent criteria, 11 healthy controls, 12 resilience to Alzheimer's disease and 20 Alzheimer's disease with dementia and analyzed matched isocortical regions, hippocampus, and caudate nucleus by mass spectrometry-based proteomics. Of 7115 differentially expressed soluble proteins, lower isocortical and hippocampal soluble Aß levels is a significant feature of resilience when compared to healthy control and Alzheimer's disease dementia groups. Protein co-expression analysis reveals 181 densely-interacting proteins significantly associated with resilience that were enriched for actin filament-based processes, cellular detoxification, and wound healing in isocortex and hippocampus, further supported by four validation cohorts. Our results suggest that lowering soluble Aß concentration may suppress severe cognitive impairment along the Alzheimer's disease continuum. The molecular basis of resilience likely holds important therapeutic insights.

A clinically applicable AI system for diagnosis of congenital heart diseases based on computed tomography images.

Congenital heart disease (CHD) is the most common type of birth defect. Without timely detection and treatment, approximately one-third of children with CHD would die in the infant period. However, due to the complicated heart structures, early diagnosis of CHD and its types is quite challenging, even for experienced radiologists. Here, we present an artificial intelligence (AI) system that achieves a comparable performance of human experts in the critical task of classifying 17 categories of CHD types. We collected the first-large CT dataset from three different CT machines, including more than 3750 CHD patients over 14 years. Experimental results demonstrate that it can achieve diagnosis accuracy (86.03%) comparable with junior cardiovascular radiologists (86.27%) in a World Health Organization appointed research and cooperation center in China on most types of CHD, and obtains a higher sensitivity (82.91%) than junior cardiovascular radiologists (76.18%). The accuracy of the combination of our AI system (97.20%) and senior radiologists achieves comparable results to that of junior radiologists and senior radiologists (97.16%) which is the current clinical routine. Our AI system can further provide 3D visualization of hearts to senior radiologists for interpretation and flexible review, surgeons for precise intuition of heart structures, and clinicians for more precise outcome prediction. We demonstrate the potential of our model to be integrated into current clinic practice to improve the diagnosis of CHD globally, especially in regions where experienced radiologists can be scarce.

Unveiling Resilience to Alzheimer's Disease: Insights From Brain Regional Proteomic Markers.

Studying proteomics data of the human brain could offer numerous insights into unraveling the signature of resilience to Alzheimer's disease. In our previous study with rigorous cohort selection criteria that excluded 4 common comorbidities, we harnessed multiple brain regions from 43 research participants with 12 of them displaying cognitive resilience to Alzheimer's disease. Based on the previous findings, this work focuses on 6 proteins out of the 33 differentially expressed proteins associated with resilience to Alzheimer's disease. These proteins are used to construct a decision tree classifier, enabling the differentiation of 3 groups: (i) healthy control, (ii) resilience to Alzheimer's disease, and (iii) Alzheimer's disease with dementia. Our analysis unveiled 2 important regional proteomic markers: Aß peptides in the hippocampus and PA1B3 in the inferior parietal lobule. These findings underscore the potential of using distinct regional proteomic markers as signatures in characterizing the resilience to Alzheimer's disease.

Canonical NF-kappaB activation is essential for Epstein-Barr virus latent membrane protein 1 TES2/CTAR2 gene regulation.

Epstein-Barr virus (EBV) latent membrane protein 1 (LMP1) transforms rodent fibroblasts and is expressed in most EBV-associated malignancies. LMP1 (transformation effector site 2 [TES2]/C-terminal activation region 2 [CTAR2]) activates NF-κB, p38, Jun N-terminal protein kinase (JNK), extracellular signal-regulated kinase (ERK), and interferon regulatory factor 7 (IRF7) pathways. We have investigated LMP1 TES2 genome-wide RNA effects at 4 time points after LMP1 TES2 expression in HEK-293 cells. By using a false discovery rate (FDR) of <0.001 after correction for multiple hypotheses, LMP1 TES2 caused >2-fold changes in 1,916 mRNAs; 1,479 RNAs were upregulated and 437 were downregulated. In contrast to tumor necrosis factor alpha (TNF-α) stimulation, which transiently upregulates many target genes, LMP1 TES2 maintained most RNA effects through the time course, despite robust and sustained induction of negative feedback regulators, such as IκBα and A20. LMP1 TES2-regulated RNAs encode many NF-κB signaling proteins and secondary interacting proteins. Consequently, many LMP1 TES2-regulated RNAs encode proteins that form an extensive interactome. Gene set enrichment analyses found LMP1 TES2-upregulated genes to be significantly enriched for pathways in cancer, B- and T-cell receptor signaling, and Toll-like receptor signaling. Surprisingly, LMP1 TES2 and IκBα superrepressor coexpression decreased LMP1 TES2 RNA effects to only 5 RNAs, with FDRs of <0.001-fold and >2-fold changes. Thus, canonical NF-κB activation is critical for almost all LMP1 TES2 RNA effects in HEK-293 cells and a more significant therapeutic target than previously appreciated.

High-Throughput Precision Phenotyping of Left Ventricular Hypertrophy With Cardiovascular Deep Learning.

IMPORTANCE: Early detection and characterization of increased left ventricular (LV) wall thickness can markedly impact patient care but is limited by under-recognition of hypertrophy, measurement error and variability, and difficulty differentiating causes of increased wall thickness, such as hypertrophy, cardiomyopathy, and cardiac amyloidosis. OBJECTIVE: To assess the accuracy of a deep learning workflow in quantifying ventricular hypertrophy and predicting the cause of increased LV wall thickness. DESIGN, SETTINGS, AND PARTICIPANTS: This cohort study included physician-curated cohorts from the Stanford Amyloid Center and Cedars-Sinai Medical Center (CSMC) Advanced Heart Disease Clinic for cardiac amyloidosis and the Stanford Center for Inherited Cardiovascular Disease and the CSMC Hypertrophic Cardiomyopathy Clinic for hypertrophic cardiomyopathy from January 1, 2008, to December 31, 2020. The deep learning algorithm was trained and tested on retrospectively obtained independent echocardiogram videos from Stanford Healthcare, CSMC, and the Unity Imaging Collaborative. MAIN OUTCOMES AND MEASURES: The main outcome was the accuracy of the deep learning algorithm in measuring left ventricular dimensions and identifying patients with increased LV wall thickness diagnosed with hypertrophic cardiomyopathy and cardiac amyloidosis. RESULTS: The study included 23 745 patients: 12 001 from Stanford Health Care (6509 [54.2%] female; mean [SD] age, 61.6 [17.4] years) and 1309 from CSMC (808 [61.7%] female; mean [SD] age, 62.8 [17.2] years) with parasternal long-axis videos and 8084 from Stanford Health Care (4201 [54.0%] female; mean [SD] age, 69.1 [16.8] years) and 2351 from CSMS (6509 [54.2%] female; mean [SD] age, 69.6 [14.7] years) with apical 4-chamber videos. The deep learning algorithm accurately measured intraventricular wall thickness (mean absolute error [MAE], 1.2 mm; 95% CI, 1.1-1.3 mm), LV diameter (MAE, 2.4 mm; 95% CI, 2.2-2.6 mm), and posterior wall thickness (MAE, 1.4 mm; 95% CI, 1.2-1.5 mm) and classified cardiac amyloidosis (area under the curve [AUC], 0.83) and hypertrophic cardiomyopathy (AUC, 0.98) separately from other causes of LV hypertrophy. In external data sets from independent domestic and international health care systems, the deep learning algorithm accurately quantified ventricular parameters (domestic: R2, 0.96; international: R2, 0.90). For the domestic data set, the MAE was 1.7 mm (95% CI, 1.6-1.8 mm) for intraventricular septum thickness, 3.8 mm (95% CI, 3.5-4.0 mm) for LV internal dimension, and 1.8 mm (95% CI, 1.7-2.0 mm) for LV posterior wall thickness. For the international data set, the MAE was 1.7 mm (95% CI, 1.5-2.0 mm) for intraventricular septum thickness, 2.9 mm (95% CI, 2.4-3.3 mm) for LV internal dimension, and 2.3 mm (95% CI, 1.9-2.7 mm) for LV posterior wall thickness. The deep learning algorithm accurately detected cardiac amyloidosis (AUC, 0.79) and hypertrophic cardiomyopathy (AUC, 0.89) in the domestic external validation site. CONCLUSIONS AND RELEVANCE: In this cohort study, the deep learning model accurately identified subtle changes in LV wall geometric measurements and the causes of hypertrophy. Unlike with human experts, the deep learning workflow is fully automated, allowing for reproducible, precise measurements, and may provide a foundation for precision diagnosis of cardiac hypertrophy.

How to evaluate deep learning for cancer diagnostics - factors and recommendations.

The large volume of data used in cancer diagnosis presents a unique opportunity for deep learning algorithms, which improve in predictive performance with increasing data. When applying deep learning to cancer diagnosis, the goal is often to learn how to classify an input sample (such as images or biomarkers) into predefined categories (such as benign or cancerous). In this article, we examine examples of how deep learning algorithms have been implemented to make predictions related to cancer diagnosis using clinical, radiological, and pathological image data. We present a systematic approach for evaluating the development and application of clinical deep learning algorithms. Based on these examples and the current state of deep learning in medicine, we discuss the future possibilities in this space and outline a roadmap for implementations of deep learning in cancer diagnosis.

Deep learning evaluation of biomarkers from echocardiogram videos.

BACKGROUND: Laboratory testing is routinely used to assay blood biomarkers to provide information on physiologic state beyond what clinicians can evaluate from interpreting medical imaging. We hypothesized that deep learning interpretation of echocardiogram videos can provide additional value in understanding disease states and can evaluate common biomarkers results. METHODS: We developed EchoNet-Labs, a video-based deep learning algorithm to detect evidence of anemia, elevated B-type natriuretic peptide (BNP), troponin I, and blood urea nitrogen (BUN), as well as values of ten additional lab tests directly from echocardiograms. We included patients (n = 39,460) aged 18 years or older with one or more apical-4-chamber echocardiogram videos (n = 70,066) from Stanford Healthcare for training and internal testing of EchoNet-Lab's performance in estimating the most proximal biomarker result. Without fine-tuning, the performance of EchoNet-Labs was further evaluated on an additional external test dataset (n = 1,301) from Cedars-Sinai Medical Center. We calculated the area under the curve (AUC) of the receiver operating characteristic curve for the internal and external test datasets. FINDINGS: On the held-out test set of Stanford patients not previously seen during model training, EchoNet-Labs achieved an AUC of 0.80 (0.79-0.81) in detecting anemia (low hemoglobin), 0.86 (0.85-0.88) in detecting elevated BNP, 0.75 (0.73-0.78) in detecting elevated troponin I, and 0.74 (0.72-0.76) in detecting elevated BUN. On the external test dataset from Cedars-Sinai, EchoNet-Labs achieved an AUC of 0.80 (0.77-0.82) in detecting anemia, of 0.82 (0.79-0.84) in detecting elevated BNP, of 0.75 (0.72-0.78) in detecting elevated troponin I, and of 0.69 (0.66-0.71) in detecting elevated BUN. We further demonstrate the utility of the model in detecting abnormalities in 10 additional lab tests. We investigate the features necessary for EchoNet-Labs to make successful detection and identify potential mechanisms for each biomarker using well-known and novel explainability techniques. INTERPRETATION: These results show that deep learning applied to diagnostic imaging can provide additional clinical value and identify phenotypic information beyond current imaging interpretation methods. FUNDING: J.W.H. and B.H. are supported by the NSF Graduate Research Fellowship. D.O. is supported by NIH K99 HL157421-01. J.Y.Z. is supported by NSF CAREER 1942926, NIH R21 MD012867-01, NIH P30AG059307 and by a Chan-Zuckerberg Biohub Fellowship.