Advances in Machine Learning Approaches to Heart Failure with Preserved Ejection Fraction.

Heart failure with preserved ejection fraction (HFpEF) represents a prototypical cardiovascular condition in which machine learning may improve targeted therapies and mechanistic understanding of pathogenesis. Machine learning, which involves algorithms that learn from data, has the potential to guide precision medicine approaches for complex clinical syndromes such as HFpEF. It is therefore important to understand the potential utility and common pitfalls of machine learning so that it can be applied and interpreted appropriately. Although machine learning holds considerable promise for HFpEF, it is subject to several potential pitfalls, which are important factors to consider when interpreting machine learning studies.

DeepCOVID-XR: An Artificial Intelligence Algorithm to Detect COVID-19 on Chest Radiographs Trained and Tested on a Large U.S. Clinical Data Set.

Background There are characteristic findings of coronavirus disease 2019 (COVID-19) on chest images. An artificial intelligence (AI) algorithm to detect COVID-19 on chest radiographs might be useful for triage or infection control within a hospital setting, but prior reports have been limited by small data sets, poor data quality, or both. Purpose To present DeepCOVID-XR, a deep learning AI algorithm to detect COVID-19 on chest radiographs, that was trained and tested on a large clinical data set. Materials and Methods DeepCOVID-XR is an ensemble of convolutional neural networks developed to detect COVID-19 on frontal chest radiographs, with reverse-transcription polymerase chain reaction test results as the reference standard. The algorithm was trained and validated on 14 788 images (4253 positive for COVID-19) from sites across the Northwestern Memorial Health Care System from February 2020 to April 2020 and was then tested on 2214 images (1192 positive for COVID-19) from a single hold-out institution. Performance of the algorithm was compared with interpretations from five experienced thoracic radiologists on 300 random test images using the McNemar test for sensitivity and specificity and the DeLong test for the area under the receiver operating characteristic curve (AUC). Results A total of 5853 patients (mean age, 58 years ± 19 [standard deviation]; 3101 women) were evaluated across data sets. For the entire test set, accuracy of DeepCOVID-XR was 83%, with an AUC of 0.90. For 300 random test images (134 positive for COVID-19), accuracy of DeepCOVID-XR was 82%, compared with that of individual radiologists (range, 76%-81%) and the consensus of all five radiologists (81%). DeepCOVID-XR had a significantly higher sensitivity (71%) than one radiologist (60%, P < .001) and significantly higher specificity (92%) than two radiologists (75%, P < .001; 84%, P = .009). AUC of DeepCOVID-XR was 0.88 compared with the consensus AUC of 0.85 (P = .13 for comparison). With consensus interpretation as the reference standard, the AUC of DeepCOVID-XR was 0.95 (95% CI: 0.92, 0.98). Conclusion DeepCOVID-XR, an artificial intelligence algorithm, detected coronavirus disease 2019 on chest radiographs with a performance similar to that of experienced thoracic radiologists in consensus. © RSNA, 2020 Supplemental material is available for this article. See also the editorial by van Ginneken in this issue.

Cases from a busy nuclear cardiology laboratory: Potential pitfalls in the interpretation of cardiac scintigraphy for ATTR cardiac amyloidosis.

Cardiac scintigraphy has emerged as a key diagnostic test for transthyretin cardiac amyloidosis (ATTR-CA). However, there are potential limitations and pitfalls in the interpretation of cardiac scintigraphy for ATTR-CA that are worth noting. We present here a series of three cases which illustrate some of these important principles.

Predicting High-Risk Patients and High-Risk Outcomes in Heart Failure.

Identifying patients with heart failure at high risk for poor outcomes is important for patient care, resource allocation, and process improvement. Although numerous risk models exist to predict mortality, hospitalization, and patient-reported health status, they are infrequently used for several reasons, including modest performance, lack of evidence to support routine clinical use, and barriers to implementation. Artificial intelligence has the potential to enhance the performance of risk prediction models, but has its own limitations and remains unproved.

Electromyographic and Force Plate Analysis of the Deadlift Performed With and Without Chains.

The purpose of this study was to determine the effects of deadlift chain variable resistance on surface electromyography (EMG) of the gluteus maximus, erector spinae, and vastus lateralis muscles, ground reaction forces (GRFs), and rate of force development (RFD). Thirteen resistance-trained men (24.0 ± 2.1 years, 179.3 ± 4.8 cm, 87.0 ± 10.6 kg) volunteered for the study. On day 1, subjects performed 1 repetition maximum (1RM) testing of the deadlift exercise. On day 2, subjects performed one set of 3 repetitions with a load of 85% 1RM with chains (CH) and without chains (NC). The order of the CH and NC conditions was randomly determined for each subject. For the CH condition, the chains accounted for approximately 20% (19.9 ± 0.6%) of the 85% 1RM load, matched at the top of the lift. Surface EMG was recorded to differentiate muscle activity between conditions (CH, NC), range of motion (ROM; bottom, top), and phase (concentric, eccentric). Peak GRFs and RFD were measured using a force plate. Electromyography results revealed that for the gluteus maximus there was significantly greater EMG activity during the NC condition vs. the CH condition. For the erector spinae, EMG activity was greater at the bottom than the top ROM (p ≤ 0.05). Force plate results revealed that deadlifting at 85% 1RM with an accommodating chain resistance of approximately 20% results in a reduction in GRFs (p ≤ 0.05) and no change in RFD (p > 0.05). Collectively, these results suggest that the use of chain resistance during deadlifting can alter muscle activation and force characteristics of the lift.

The interplay between interface structure, energy level alignment and chemical bonding strength at organic-metal interfaces.

What do energy level alignments at metal-organic interfaces reveal about the metal-molecule bonding strength? Is it permissible to take vertical adsorption heights as indicators of bonding strengths? In this paper we analyse 3,4,9,10-perylene-tetracarboxylic acid dianhydride (PTCDA) on the three canonical low index Ag surfaces to provide exemplary answers to these questions. Specifically, we employ angular resolved photoemission spectroscopy for a systematic study of the energy level alignments of the two uppermost frontier states in ordered monolayer phases of PTCDA. Data are analysed using the orbital tomography approach. This allows the unambiguous identification of the orbital character of these states, and also the discrimination between inequivalent species. Combining this experimental information with DFT calculations and the generic Newns-Anderson chemisorption model, we analyse the alignments of highest occupied and lowest unoccupied molecular orbitals (HOMO and LUMO) with respect to the vacuum levels of bare and molecule-covered surfaces. This reveals clear differences between the two frontier states. In particular, on all surfaces the LUMO is subject to considerable bond stabilization through the interaction between the molecular π-electron system and the metal, as a consequence of which it also becomes occupied. Moreover, we observe a larger bond stabilization for the more open surfaces. Most importantly, our analysis shows that both the orbital binding energies of the LUMO and the overall adsorption heights of the molecule are linked to the strength of the chemical interaction between the molecular π-electron system and the metal, in the sense that stronger bonding leads to shorter adsorption heights and larger orbital binding energies.

Repeated change-of-direction test for collegiate male soccer players.

AIM: The aim of the study was to investigate the applicability of a repeated change-of-direction (RCoD) test for NCAA Division-I male soccer players. METHODS: The RCoD test consisted of 5 diagonal direction changes per repetition with a soccer ball to be struck at the end. Each player performed 15 repetitions with approximately 10 seconds to jog back between repetitions. Data were collected in two sessions. In the first session, 13 players were examined for heart rate responses and blood lactate concentrations. In the second session, 22 players were examined for the test's ability to discriminate the primary from secondary players (78.0±16.1 and 10.4±13.3 minutes per match, respectively). RESULTS: Heart rate data were available only from 9 players due to artifacts. The peak heart rate (200.2±6.6 beats∙min-1: 99.9±3.0% maximum) and blood lactate concentration (14.8±2.4 mmol∙L-1 immediately after) resulted in approximately 3.5 and 6.4-fold increases from the resting values, respectively. These values appear comparable to those during intense periods of soccer matches. In addition, the average repetition time of the test was found to discriminate the primary (4.85±0.23 s) from the secondary players (5.10±0.24 s) (P=0.02). CONCLUSION: The RCoD test appears to induce physiological responses similar to intense periods of soccer matches with respect to heart rate and blood lactate concentration. Players with better average repetition times tend to be those who play major minutes.

Investigating the stretch-shortening cycle fatigue response to a high-intensity stressful phase of training in collegiate men's basketball.

Introduction: While using force-plate derived measures of vertical jump performance, reflective of stretch-shortening-cycle (SSC) efficiency is common practice in sport science, there is limited evidence as to which tests and measures may be most sensitive toward neuromuscular fatigue. The aim of this study was to explore the SSC fatigue response to a one-week high-intensity fatiguing phase of training in National Collegiate Athletic Association (NCAA) Division-I basketball players. Methods: The study timeline consisted of three weeks of baseline measures, one week of high-intensity training, and two weeks of follow-up testing. Countermovement jumps (CMJ) and 10-5 hop tests were performed at baseline, as well as at two time-points during, and three time-points following the fatiguing training period, allowing for performance-comparisons with baseline. Results: Compared to the weekly training sum at baseline, during the high intensity training phase, athletes were exposed to very large increases in selected external load metrics (ES = 1.44-3.16), suggesting that athletes experienced fatigue acutely, as well as potential longer lasting reductions in performance. Vertical jump data suggested that in the CMJ, traditional metrics such as jump height, as well as metrics reflecting kinetic outputs and movement strategies, were sensitive to the stark increase in high-intensity training exposure. The 10-5 hop test suggested a fatigue-induced loss of tolerance to ground impact reflected by performance reductions in metrics related to jump height and reactive strength qualities. Discussion: These findings emphasize that when monitoring neuromuscular fatigue, variables and assessments may not be looked at individually, but rather as part of a more global monitoring approach.

Accuracy and consistency of online large language model-based artificial intelligence chat platforms in answering patients' questions about heart failure.

BACKGROUND: Heart failure (HF) is a prevalent condition associated with significant morbidity. Patients may have questions that they feel embarrassed to ask or will face delays awaiting responses from their healthcare providers which may impact their health behavior. We aimed to investigate the potential of large language model (LLM) based artificial intelligence (AI) chat platforms in complementing the delivery of patient-centered care. METHODS: Using online patient forums and physician experience, we created 30 questions related to diagnosis, management and prognosis of HF. The questions were posed to two LLM-based AI chat platforms (OpenAI's ChatGPT-3.5 and Google's Bard). Each set of answers was evaluated by two HF experts, independently and blinded to each other, for accuracy (adequacy of content) and consistency of content. RESULTS: ChatGPT provided mostly appropriate answers (27/30, 90%) and showed a high degree of consistency (93%). Bard provided a similar content in its answers and thus was evaluated only for adequacy (23/30, 77%). The two HF experts' grades were concordant in 83% and 67% of the questions for ChatGPT and Bard, respectively. CONCLUSION: LLM-based AI chat platforms demonstrate potential in improving HF education and empowering patients, however, these platforms currently suffer from issues related to factual errors and difficulty with more contemporary recommendations. This inaccurate information may pose serious and life-threatening implications for patients that should be considered and addressed in future research.

Performance of risk models to predict mortality risk for patients with heart failure: evaluation in an integrated health system.

BACKGROUND: Referral of patients with heart failure (HF) who are at high mortality risk for specialist evaluation is recommended. Yet, most tools for identifying such patients are difficult to implement in electronic health record (EHR) systems. OBJECTIVE: To assess the performance and ease of implementation of Machine learning Assessment of RisK and EaRly mortality in Heart Failure (MARKER-HF), a machine-learning model that uses structured data that is readily available in the EHR, and compare it with two commonly used risk scores: the Seattle Heart Failure Model (SHFM) and Meta-Analysis Global Group in Chronic (MAGGIC) Heart Failure Risk Score. DESIGN: Retrospective, cohort study. PARTICIPANTS: Data from 6764 adults with HF were abstracted from EHRs at a large integrated health system from 1/1/10 to 12/31/19. MAIN MEASURES: One-year survival from time of first cardiology or primary care visit was estimated using MARKER-HF, SHFM, and MAGGIC. Discrimination was measured by the area under the receiver operating curve (AUC). Calibration was assessed graphically. KEY RESULTS: Compared to MARKER-HF, both SHFM and MAGGIC required a considerably larger amount of data engineering and imputation to generate risk score estimates. MARKER-HF, SHFM, and MAGGIC exhibited similar discriminations with AUCs of 0.70 (0.69-0.73), 0.71 (0.69-0.72), and 0.71 (95% CI 0.70-0.73), respectively. All three scores showed good calibration across the full risk spectrum. CONCLUSIONS: These findings suggest that MARKER-HF, which uses readily available clinical and lab measurements in the EHR and required less imputation and data engineering than SHFM and MAGGIC, is an easier tool to identify high-risk patients in ambulatory clinics who could benefit from referral to a HF specialist.

The assessment of vascular risk in men with erectile dysfunction: the role of the cardiologist and general physician.

Erectile dysfunction (ED) and cardiovascular disease (CVD) share risk factors and frequently coexist, with endothelial dysfunction believed to be the pathophysiologic link. ED is common, affecting more than 70% of men with known CVD. In addition, clinical studies have demonstrated that ED in men with no known CVD often precedes a CVD event by 2-5 years. ED severity has been correlated with increasing plaque burden in patients with coronary artery disease. ED is an independent marker of increased CVD risk including all-cause and especially CVD mortality, particularly in men aged 30-60 years. Thus, ED identifies a window of opportunity for CVD risk mitigation. We recommend that a thorough history, physical exam (including visceral adiposity), assessment of ED severity and duration and evaluation including fasting plasma glucose, lipids, resting electrocardiogram, family history, lifestyle factors, serum creatinine (estimated glomerular filtration rate) and albumin:creatinine ratio, and determination of the presence or absence of the metabolic syndrome be performed to characterise cardiovascular risk in all men with ED. Assessment of testosterone levels should also be considered and biomarkers may help to further quantify risk, even though their roles in development of CVD have not been firmly established. Finally, we recommend that a question about ED be included in assessment of CVD risk in all men and be added to CVD risk assessment guidelines.

Relationships of isometric mid-thigh pull variables to weightlifting performance.

AIM: The purpose of this study was to evaluate the relationship between weightlifting performance (snatch, clean and jerk, and total) and variables obtained from the isometric mid-thigh pull (IMTP). METHODS: Twelve weightlifters, ranging from novice to advanced, performed the IMTP 10 days after a competition. Correlations were used to evaluate relationships between variables of the IMTP and absolute and scaled competition results. RESULTS: Unscaled competition results correlated strongly with IRFD (0-200ms: r=0.567-0.645, 0-250ms: r=0.722-0.781) while results correlated weakly with Peak IRFD (5ms window, r=0.360-0.426). Absolute peak force values correlated very strongly with absolute values for the competition performance (r=0.830-0.838). Force at 100ms, 150ms, 200ms and 250ms also correlated strongly with competition results (r=0.643-0.647, r=0.605-0.636, r=0.714-0.732, r=0.801-0.804). Similar findings were noted for allometrically scaled values. CONCLUSION: Measures of average IRFD probably represent a more relevant variable to dynamic performance than does Peak IRFD (5ms). Maximum isometric strength also is likely to have a strong role in weightlifting performance.

A comparative study of pretrained language models for long clinical text.

OBJECTIVE: Clinical knowledge-enriched transformer models (eg, ClinicalBERT) have state-of-the-art results on clinical natural language processing (NLP) tasks. One of the core limitations of these transformer models is the substantial memory consumption due to their full self-attention mechanism, which leads to the performance degradation in long clinical texts. To overcome this, we propose to leverage long-sequence transformer models (eg, Longformer and BigBird), which extend the maximum input sequence length from 512 to 4096, to enhance the ability to model long-term dependencies in long clinical texts. MATERIALS AND METHODS: Inspired by the success of long-sequence transformer models and the fact that clinical notes are mostly long, we introduce 2 domain-enriched language models, Clinical-Longformer and Clinical-BigBird, which are pretrained on a large-scale clinical corpus. We evaluate both language models using 10 baseline tasks including named entity recognition, question answering, natural language inference, and document classification tasks. RESULTS: The results demonstrate that Clinical-Longformer and Clinical-BigBird consistently and significantly outperform ClinicalBERT and other short-sequence transformers in all 10 downstream tasks and achieve new state-of-the-art results. DISCUSSION: Our pretrained language models provide the bedrock for clinical NLP using long texts. We have made our source code available at https://github.com/luoyuanlab/Clinical-Longformer, and the pretrained models available for public download at: https://huggingface.co/yikuan8/Clinical-Longformer. CONCLUSION: This study demonstrates that clinical knowledge-enriched long-sequence transformers are able to learn long-term dependencies in long clinical text. Our methods can also inspire the development of other domain-enriched long-sequence transformers.

Changes in countermovement jump force-time characteristic in elite male basketball players: A season-long analyses.

Basketball is a sport that is characterized by various physical performance parameters and motor abilities such as speed, strength, and endurance, which are all underpinned by an athlete's efficient use of the stretch-shortening cycle (SSC). A common assessment to measure SSC efficiency is the countermovement jump (CMJ). When performed on a force plate, a plethora of different force-time metrics may be gleaned from the jump task, reflecting neuromuscular performance characteristics. The aim of this study was to investigate how different CMJ force-time characteristics change across different parts of the athletic year, within a sample of elite collegiate male basketball players. Twelve basketball players performed CMJ's on near-weekly basis, combining for a total of 219 screenings. The span of testing was broken down into four periods: pre-season, non-conference competitive period, conference competitive period, and post-season competitive period. Results suggest that basketball players were able to experience improvements and maintenance of performance with regards to various force-time metrics, transitioning from the pre-season period into respective later phases of the in-season period. A common theme was a significant improvement between the pre-season period and the non-conference period. Various force-time metrics were subject to change, while outcome metrics such as jump height remained unchanged, suggesting that practitioners are encouraged to more closely monitor how different force-time characteristics change over extended periods of time.

Vertical Jump Neuromuscular Performance Characteristics Determining On-Court Contribution in Male and Female NCAA Division 1 Basketball Players.

While various quantifiable physical attributes have been found to contribute to athletes' performance, there is a lack of scientific literature focused on examining how they relate to success during competition performance. The aim of this study was to investigate different countermovement jump (CMJ)-derived force-time characteristics and their utility in distinguishing high from low performers within a measure of on-court contribution (i.e., minutes per game played). Twenty-nine collegiate athletes (n = 15 males and n = 14 females) volunteered to participate in this investigation and performed CMJs on dual force plates sampling at 1000 Hz, weekly over the course of their basketball season. The athletes' average of their three best test-days across the season was used for further analysis. To identify their on-court contribution, athletes were divided into groups with high and low minutes per game, based on a median-split analysis. The findings suggest that at the overall group level (i.e., both genders), the modified reactive strength index (mRSI) and braking rate of force development (RFD) revealed the greatest between-group magnitudes of difference, with athletes playing more minutes per game showing greater performance. At the team-specific level, the braking RFD, average braking velocity, and mRSI were shown to be the greatest differentiators between groups for the men's team. The women's high-minutes group displayed greater magnitudes of mRSI and jump height. By identifying the neuromuscular qualities seen in top performers within their respective populations, the attributed physical performance underpinning these qualities may be identified, providing practitioners with insights into physical performance qualities and training methodologies that have the potential to influence basketball performance.

Microporogen-Structured Collagen Matrices for Embedded Bioprinting of Tumor Models for Immuno-Oncology.

Embedded bioprinting enables the rapid design and fabrication of complex tissues that recapitulate in vivo microenvironments. However, few biological matrices enable good print fidelity, while simultaneously facilitate cell viability, proliferation, and migration. Here, a new microporogen-structured (µPOROS) matrix for embedded bioprinting is introduced, in which matrix rheology, printing behavior, and porosity are tailored by adding sacrificial microparticles composed of a gelatin-chitosan complex to a prepolymer collagen solution. To demonstrate its utility, a 3D tumor model is created via embedded printing of a murine melanoma cell ink within the µPOROS collagen matrix at 4 °C. The collagen matrix is subsequently crosslinked around the microparticles upon warming to 21 °C, followed by their melting and removal at 37 °C. This process results in a µPOROS matrix with a fibrillar collagen type-I network akin to that observed in vivo. Printed tumor cells remain viable and proliferate, while antigen-specific cytotoxic T cells incorporated in the matrix migrate to the tumor site, where they induce cell death. The integration of the µPOROS matrix with embedded bioprinting opens new avenues for creating complex tissue microenvironments in vitro that may find widespread use in drug discovery, disease modeling, and tissue engineering for therapeutic use.

Deep Learning for Cardiovascular Imaging: A Review.

Importance: Artificial intelligence (AI), driven by advances in deep learning (DL), has the potential to reshape the field of cardiovascular imaging (CVI). While DL for CVI is still in its infancy, research is accelerating to aid in the acquisition, processing, and/or interpretation of CVI across various modalities, with several commercial products already in clinical use. It is imperative that cardiovascular imagers are familiar with DL systems, including a basic understanding of how they work, their relative strengths compared with other automated systems, and possible pitfalls in their implementation. The goal of this article is to review the methodology and application of DL to CVI in a simple, digestible fashion toward demystifying this emerging technology. Observations: At its core, DL is simply the application of a series of tunable mathematical operations that translate input data into a desired output. Based on artificial neural networks that are inspired by the human nervous system, there are several types of DL architectures suited to different tasks; convolutional neural networks are particularly adept at extracting valuable information from CVI data. We survey some of the notable applications of DL to tasks across the spectrum of CVI modalities. We also discuss challenges in the development and implementation of DL systems, including avoiding overfitting, preventing systematic bias, improving explainability, and fostering a human-machine partnership. Finally, we conclude with a vision of the future of DL for CVI. Conclusions and Relevance: Deep learning has the potential to meaningfully affect the field of CVI. Rather than a threat, DL could be seen as a partner to cardiovascular imagers in reducing technical burden and improving efficiency and quality of care. High-quality prospective evidence is still needed to demonstrate how the benefits of DL CVI systems may outweigh the risks.

Noncanonical role of singleminded-2s in mitochondrial respiratory chain formation in breast cancer.

Dysregulation of cellular metabolism is a hallmark of breast cancer progression and is associated with metastasis and therapeutic resistance. Here, we show that the breast tumor suppressor gene SIM2 promotes mitochondrial oxidative phosphorylation (OXPHOS) using breast cancer cell line models. Mechanistically, we found that SIM2s functions not as a transcription factor but localizes to mitochondria and directly interacts with the mitochondrial respiratory chain (MRC) to facilitate functional supercomplex (SC) formation. Loss of SIM2s expression disrupts SC formation through destabilization of MRC Complex III, leading to inhibition of electron transport, although Complex I (CI) activity is retained. A metabolomic analysis showed that knockout of SIM2s leads to a compensatory increase in ATP production through glycolysis and accelerated glutamine-driven TCA cycle production of NADH, creating a favorable environment for high cell proliferation. Our findings indicate that SIM2s is a novel stabilizing factor required for SC assembly, providing insight into the impact of the MRC on metabolic adaptation and breast cancer progression.

Analysis of perioperative outcomes, length of hospital stay, and readmission rate after gastric bypass.

BACKGROUND: Hospital lengths of stay (LOS) and readmission rates often are used by third parties to measure quality of outcomes despite only a few published series that analyze risk-adjusted data and predictors of these events. METHODS: Single-institution retrospective multivariable analysis of consecutive Roux-en-Y gastric bypass (RYGB) patients was performed to determine variables that may influence LOS and the readmission rate. RESULTS: Between 2006 and 2010, 1,065 consecutive RYGB procedures were analyzed. The mean initial body mass index (BMI) of the patients was 48.4 kg/m(2) (range 35-108 kg/m(2)), and their mean age was 42 years (range 15-75 years). Of these patients, 42% were black and 31% were either Medicare or Medicaid beneficiaries. The average LOS was 1.8 days (range 1-59 days; median, 2 days). The hospital discharged 48% of these patients on postoperative day (POD) 1, 85% on POD 2, and 96% on POD 3. According to multivariable Poisson regression, the independent predictors of a longer LOS included longer procedure time, surgeon, BMI, black race, older age, and status as a Medicare/Medicaid beneficiary (all P < 0.01). Gender and measured comorbidities were not associated with LOS. However, this model was poorly predictive of LOS due to substantial unexplained variance (R (2) = 0.10). Complications were significantly associated with Medicare/Medicare status (odds ratio [OR] 2.0), older age (OR 1.03), and longer procedure time (OR 1.02) (P < 0.05). According to logistic regression, a 30-day readmission rate was predicted only by a LOS longer than 3 days for the primary procedure (P < 0.0005). CONCLUSIONS: Early discharge on postoperative day 1 is possible but nonmodifiable, and random patient factors challenge predictable discharge planning. Reliable discharge on postoperative day 1 is not likely with current technologies.