Enhancing Type 2 Diabetes Treatment Decisions With Interpretable Machine Learning Models for Predicting Hemoglobin A1c Changes: Machine Learning Model Development.

BACKGROUND: Type 2 diabetes (T2D) is a significant global health challenge. Physicians need to assess whether future glycemic control will be poor on the current trajectory of usual care and usual-care treatment intensifications so that they can consider taking extra treatment measures to prevent poor outcomes. Predicting poor glycemic control from trends in hemoglobin A1c (HbA1c) levels is difficult due to the influence of seasonal fluctuations and other factors. OBJECTIVE: We sought to develop a model that accurately predicts poor glycemic control among patients with T2D receiving usual care. METHODS: Our machine learning model predicts poor glycemic control (HbA1c≥8%) using the transformer architecture, incorporating an attention mechanism to process irregularly spaced HbA1c time series and quantify temporal relationships of past HbA1c levels at each time point. We assessed the model using HbA1c levels from 7787 patients with T2D seeing specialist physicians at the University of Tokyo Hospital. The training data include instances of poor glycemic control occurring during usual care with usual-care treatment intensifications. We compared prediction accuracy, assessed with the area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate, to that of LightGBM. RESULTS: The area under the receiver operating characteristic curve, the area under the precision-recall curve, and the accuracy rate (95% confidence limits) of the proposed model were 0.925 (95% CI 0.923-0.928), 0.864 (95% CI 0.852-0.875), and 0.864 (95% CI 0.86-0.869), respectively. The proposed model achieved high prediction accuracy comparable to or surpassing LightGBM's performance. The model prioritized the most recent HbA1c levels for predictions. Older HbA1c levels in patients with poor glycemic control were slightly more influential in predictions compared to patients with good glycemic control. CONCLUSIONS: The proposed model accurately predicts poor glycemic control for patients with T2D receiving usual care, including patients receiving usual-care treatment intensifications, allowing physicians to identify cases warranting extraordinary treatment intensifications. If used by a nonspecialist, the model's indication of likely future poor glycemic control may warrant a referral to a specialist. Future efforts could incorporate diverse and large-scale clinical data for improved accuracy.

Association between pupillary examinations and prognosis in patients with out-of-hospital cardiac arrest who underwent extracorporeal cardiopulmonary resuscitation: a retrospective multicentre cohort study.

BACKGROUND: In some cases of patients with out-of-hospital cardiac arrest (OHCA) who underwent extracorporeal cardiopulmonary resuscitation (ECPR), negative pupillary light reflex (PLR) and mydriasis upon hospital arrival serve as common early indicator of poor prognosis. However, in certain patients with poor prognoses inferred by pupil findings upon hospital arrival, pupillary findings improve before and after the establishment of ECPR. The association between these changes in pupillary findings and prognosis remains unclear. This study aimed to clarify the association of pupillary examinations before and after the establishment of ECPR in patients with OHCA showing poor pupillary findings upon hospital arrival with their outcomes. To this end, we analysed retrospective multicentre registry data involving 36 institutions in Japan, including all adult patients with OHCA who underwent ECPR between January 2013 and December 2018. We selected patients with poor prognosis inferred by pupillary examinations, negative pupillary light reflex (PLR) and pupil mydriasis, upon hospital arrival. The primary outcome was favourable neurological outcome, defined as Cerebral Performance Category 1 or 2 at hospital discharge. Multivariable logistic regression analysis was performed to evaluate the association between favourable neurological outcome and pupillary examination after establishing ECPR. RESULTS: Out of the 2,157 patients enrolled in the SAVE-J II study, 723 were analysed. Among the patients analysed, 74 (10.2%) demonstrated favourable neurological outcome at hospital discharge. Multivariable analysis revealed that a positive PLR at ICU admission (odds ration [OR] = 11.3, 95% confidence intervals [CI] = 5.17-24.7) was significantly associated with favourable neurological outcome. However, normal pupil diameter at ICU admission (OR = 1.10, 95%CI = 0.52-2.32) was not significantly associated with favourable neurological outcome. CONCLUSION: Among the patients with OHCA who underwent ECPR and showed poor pupillary examination findings upon hospital arrival, 10.2% had favourable neurological outcome at hospital discharge. A positive PLR after the establishment of ECPR was significantly associated with favourable neurological outcome.

Immediate postnatal prediction of death or bronchopulmonary dysplasia among very preterm and very low birth weight infants based on gradient boosting decision trees algorithm: A nationwide database study in Japan.

INTRODUCTION: Bronchopulmonary dysplasia (BPD) poses a substantial global health burden. Individualized treatment strategies based on early prediction of the development of BPD can mitigate preterm birth complications; however, previously suggested predictive models lack early postnatal applicability. We aimed to develop predictive models for BPD and mortality based on immediate postnatal clinical data. METHODS: Clinical information on very preterm and very low birth weight infants born between 2008 and 2018 was extracted from a nationwide Japanese database. The gradient boosting decision trees (GBDT) algorithm was adopted to predict BPD and mortality, using predictors within the first 6 h postpartum. We assessed the temporal validity and evaluated model adequacy using Shapley additive explanations (SHAP) values. RESULTS: We developed three predictive models using data from 39,488, 39,096, and 40,291 infants to predict "death or BPD," "death or severe BPD," and "death before discharge," respectively. These well-calibrated models achieved areas under the receiver operating characteristic curve of 0.828 (95% CI: 0.828-0.828), 0.873 (0.873-0.873), and 0.887 (0.887-0.888), respectively, outperforming the multivariable logistic regression models. SHAP value analysis identified predictors of BPD, including gestational age, size at birth, male sex, and persistent pulmonary hypertension. In SHAP value-based case clustering, the "death or BPD" prediction model stratified infants by gestational age and persistent pulmonary hypertension, whereas the other models for "death or severe BPD" and "death before discharge" commonly formed clusters of low mortality, extreme prematurity, low Apgar scores, and persistent pulmonary hypertension of the newborn. CONCLUSIONS: GBDT models for predicting BPD and mortality, designed for use within 6 h postpartum, demonstrated superior prognostic performance. SHAP value-based clustering, a data-driven approach, formed clusters of clinical relevance. These findings suggest the efficacy of a GBDT algorithm for the early postnatal prediction of BPD.

Factors associated with favourable neurological outcomes following cardiopulmonary resuscitation for out-of-hospital cardiac arrest: A retrospective multi-centre cohort study.

Aim: To investigate the factors associated with favourable neurological outcomes in adult patients undergoing extracorporeal cardiopulmonary resuscitation (ECPR) for out-of-hospital cardiac arrest (OHCA). Methods: This retrospective observational study used secondary analysis of the SAVE-J II multicentre registry data from 36 institutions in Japan. Between 2013 and 2018, 2157 patients with OHCA who underwent ECPR were enrolled in SAVE-J II. A total of 1823 patients met the study inclusion criteria. Adult patients (aged ≥ 18 years) with OHCA, who underwent ECPR before admission to the intensive care unit, were included in our secondary analysis. The primary outcome was a favourable neurological outcome at hospital discharge, defined as a Cerebral Performance Category score of 1 or 2. We used a multivariate logistic regression model to examine the association between factors measured at the incident scene or upon hospital arrival and favourable neurological outcomes. Results: Multivariable analysis revealed that shockable rhythm at the scene [odds ratio (OR); 2.11; 95% confidence interval (CI), 1.16-3.95] and upon hospital arrival (OR 2.59; 95% CI 1.60-4.30), bystander CPR (OR 1.63; 95% CI 1.03-1.88), body movement during resuscitation (OR 7.10; 95% CI 1.79-32.90), gasping (OR 4.33; 95% CI 2.57-7.28), pupillary reflex on arrival (OR 2.93; 95% CI 1.73-4.95), and male sex (OR 0.43; 95% CI 0.24-0.75) significantly correlated with neurological outcomes. Conclusions: Shockable rhythm, bystander CPR, body movement during resuscitation, gasping, pupillary reflex, and sex were associated with favourable neurological outcomes in patients with OHCA treated with ECPR.

Providing Practical Knowledge and Skills to Handle Real-World Data? Lessons Learned from Med RWD Program.

Both lectures and hands-on education are essential for the development of human resources that can use real-world data (RWD). The University of Tokyo has launched a new hybrid-style RWD educational program entitled "Medical Real World Data Utilization Human Resource Development Project" from FY2019 onwards. We present an overview of the overall picture of the project, including the development process of the educational program and the challenges associated with it.

Graph Representation Learning-Based Fixed-Length Clinical Feature Vector Generation from Heterogeneous Medical Records.

Transformation of patient data extracted from a database into fixed-length numerical vectors requires expertise in topical medical knowledge as well as data manipulation-thus, manual feature design is labor-intensive. In this study, we propose a machine learning-based method to for this purpose applicable to electronic medical data recorded during hospitalization, which utilizes unsupervised feature extraction based on graph embedding. Unsupervised learning is performed on a heterogeneous graph using Graph2Vec, and the inclusion of clinically useful data in the obtained embedding representation is evaluated by predicting readmission within 30 days of discharge based on it. The embedded representations are observed to improve predictive performance significantly as the information contained in the graph increases, indicating the suitability of the proposed method for feature design corresponding to clinical information.

Etiology-Based Prognosis of Extracorporeal CPR Recipients After Out-of-Hospital Cardiac Arrest: A Retrospective Multicenter Cohort Study.

BACKGROUND: A better understanding of the relative contributions of various factors to patient outcomes is essential for optimal patient selection for extracorporeal CPR (ECPR) therapy for patients with out-of-hospital cardiac arrest (OHCA). However, evidence on the prognostic comparison based on the etiologies of cardiac arrest is limited. RESEARCH QUESTION: What is the etiology-based prognosis of patients undergoing ECPR for OHCA? STUDY DESIGN AND METHODS: This retrospective multicenter registry study involved 36 institutions in Japan and included all adult patients with OHCA who underwent ECPR between January 2013 and December 2018. The primary etiology for OHCA was determined retrospectively from all hospital-based data at each institution. We performed a multivariable logistic regression model to determine the association between etiology of cardiac arrest and two outcomes: favorable neurologic outcome and survival at hospital discharge. RESULTS: We identified 1,781 eligible patients, of whom 1,405 (78.9%) had cardiac arrest because of cardiac causes. Multivariable logistic regression analysis for favorable neurologic outcome showed that accidental hypothermia (adjusted OR, 5.12; 95% CI, 2.98-8.80; P < .001) was associated with a significantly higher rate of favorable neurologic outcome than cardiac causes. Multivariable logistic regression analysis for survival showed that accidental hypothermia (adjusted OR, 5.19; 95% CI, 3.15-8.56; P < .001) had significantly higher rates of survival than cardiac causes. Acute aortic dissection/aneurysm (adjusted OR, 0.07; 95% CI, 0.02-0.28; P < .001) and primary cerebral disorders (adjusted OR, 0.12; 95% CI, 0.03-0.50; P = .004) had significantly lower rates of survival than cardiac causes. INTERPRETATION: In this retrospective multicenter cohort study, although most patients with OHCA underwent ECPR for cardiac causes, accidental hypothermia was associated with favorable neurologic outcome and survival; in contrast, acute aortic dissection/aneurysm and primary cerebral disorders were associated with nonsurvival compared with cardiac causes.

Sex- and Age-Based Disparities in Public Access Defibrillation, Bystander Cardiopulmonary Resuscitation, and Neurological Outcome in Cardiac Arrest.

Importance: Younger females with out-of-hospital cardiac arrest (OHCA) in public locations have less likelihood to receive public access defibrillation and bystander cardiopulmonary resuscitation (CPR). However, the association between age- and sex-based disparities and neurological outcomes remains underexamined. Objective: To investigate the association between sex and age and the rate of receiving bystander CPR, automated external defibrillator defibrillation, and neurological outcomes in patients with OHCA. Design, Setting, and Participants: This cohort study used the All-Japan Utstein Registry, a prospective, population-based, nationwide database in Japan containing data on 1 930 273 patients with OHCA between January 1, 2005, and December 31, 2020. The cohort comprised patients with OHCA of cardiac origin that was witnessed by citizens and treated by emergency medical service personnel. The data were analyzed from September 3, 2022, to May 5, 2023. Exposure: Sex and age. Main Outcomes and Measures: The primary outcome was favorable neurological outcome at 30 days after an OHCA. Favorable neurological outcome was defined as a Cerebral Performance Category score of 1 (indicating good cerebral performance) or 2 (indicating moderate cerebral disability). The secondary outcomes were the rates of receiving public access defibrillation and bystander CPR. Results: The 354 409 included patients who experienced bystander-witnessed OHCA of cardiac origin had a median (IQR) age of 78 (67-86) years and 136 520 were females (38.5%). The rate of receiving public access defibrillation was higher in males than females (3.2% vs 1.5%; P < .001). Stratified by age, age- and sex-based disparities in prehospital lifesaving interventions by bystanders and in neurological outcomes were observed. Although younger females had a lower rate of receiving public access defibrillation and bystander CPR than males, younger females had a higher favorable neurological outcome compared with males of the same age (odds ratio [OR], 1.19; 95% CI, 1.08-1.31). In younger females with OHCA that was witnessed by nonfamily bystanders, receiving public access defibrillation (OR, 3.51; 95% CI, 2.34-5.27) or bystander CPR (OR, 1.62; 95% CI, 1.20-2.22) was associated with a favorable neurological outcome. Conclusions and Relevance: Results of this study suggest a pattern of significant sex- and age-based differences in bystander CPR, public access defibrillation, and neurological outcomes in Japan. Improvement in neurological outcomes in patients with OHCA, especially younger females, was associated with increased use of public access defibrillation and bystander CPR.

Clinical Feature Vector Generation using Unsupervised Graph Representation Learning from Heterogeneous Medical Records.

The diversity of patient information recorded on electronic medical records generally, presents a challenge for converting it into fixed-length vectors that align with clinical characteristics. To address this issue, this study aimed to utilize an unsupervised graph representation learning method to transform the unstructured inpatient information from electronic medical records into a fixed-length vector. Infograph, one of the unsupervised graph representation learning algorithms was applied to the graphed inpatient information, resulting in embedded vectors of fixed length. The embedded vectors were then evaluated for whether the clinical information was preserved in it. The results indicated that the embedded representation contained information that could predict readmission within 30 days, demonstrating the feasibility of using unsupervised graph representation learning to transform patient information into fixed-length vectors that retain clinical characteristics.

Development of non-bias phenotypic drug screening for cardiomyocyte hypertrophy by image segmentation using deep learning.

The number of patients with heart failure and related deaths is rapidly increasing worldwide, making it a major problem. Cardiac hypertrophy is a crucial preliminary step in heart failure, but its treatment has not yet been fully successful. In this study, we established a system to evaluate cardiomyocyte hypertrophy using a deep learning-based high-throughput screening system and identified drugs that inhibit it. First, primary cultured cardiomyocytes from neonatal rats were stimulated by both angiotensin II and endothelin-1, and cellular images were captured using a phase-contrast microscope. Subsequently, we used a deep learning model for instance segmentation and established a system to automatically and unbiasedly evaluate the cardiomyocyte size and perimeter. Using this system, we screened 100 FDA-approved drugs library and identified 12 drugs that inhibited cardiomyocyte hypertrophy. We focused on ezetimibe, a cholesterol absorption inhibitor, that inhibited cardiomyocyte hypertrophy in a dose-dependent manner in vitro. Additionally, ezetimibe improved the cardiac dysfunction induced by pressure overload in mice. These results suggest that the deep learning-based system is useful for the evaluation of cardiomyocyte hypertrophy and drug screening, leading to the development of new treatments for heart failure.

The complement C3-complement factor D-C3a receptor signalling axis regulates cardiac remodelling in right ventricular failure.

Failure of the right ventricle plays a critical role in any type of heart failure. However, the mechanism remains unclear, and there is no specific therapy. Here, we show that the right ventricle predominantly expresses alternative complement pathway-related genes, including Cfd and C3aR1. Complement 3 (C3)-knockout attenuates right ventricular dysfunction and fibrosis in a mouse model of right ventricular failure. C3a is produced from C3 by the C3 convertase complex, which includes the essential component complement factor D (Cfd). Cfd-knockout mice also show attenuation of right ventricular failure. Moreover, the plasma concentration of CFD correlates with the severity of right ventricular failure in patients with chronic right ventricular failure. A C3a receptor (C3aR) antagonist dramatically improves right ventricular dysfunction in mice. In summary, we demonstrate the crucial role of the C3-Cfd-C3aR axis in right ventricular failure and highlight potential therapeutic targets for right ventricular failure.

Pathway importance by graph convolutional network and Shapley additive explanations in gene expression phenotype of diffuse large B-cell lymphoma.

Deep learning techniques have recently been applied to analyze associations between gene expression data and disease phenotypes. However, there are concerns regarding the black box problem: it is difficult to interpret why the prediction results are obtained using deep learning models from model parameters. New methods have been proposed for interpreting deep learning model predictions but have not been applied to genetics. In this study, we demonstrated that applying SHapley Additive exPlanations (SHAP) to a deep learning model using graph convolutions of genetic pathways can provide pathway-level feature importance for classification prediction of diffuse large B-cell lymphoma (DLBCL) gene expression subtypes. Using Kyoto Encyclopedia of Genes and Genomes pathways, a graph convolutional network (GCN) model was implemented to construct graphs with nodes and edges. DLBCL datasets, including microarray gene expression data and clinical information on subtypes (germinal center B-cell-like type and activated B-cell-like type), were retrieved from the Gene Expression Omnibus to evaluate the model. The GCN model showed an accuracy of 0.914, precision of 0.948, recall of 0.868, and F1 score of 0.906 in analysis of the classification performance for the test datasets. The pathways with high feature importance by SHAP included highly enriched pathways in the gene set enrichment analysis. Moreover, a logistic regression model with explanatory variables of genes in pathways with high feature importance showed good performance in predicting DLBCL subtypes. In conclusion, our GCN model for classifying DLBCL subtypes is useful for interpreting important regulatory pathways that contribute to the prediction.

Deep learning model to detect significant aortic regurgitation using electrocardiography.

BACKGROUND: Aortic regurgitation (AR) is a common heart disease, with a relatively high prevalence of 4.9% in the Framingham Heart Study. Because the prevalence increases with advancing age, an upward shift in the age distribution may increase the burden of AR. To provide an effective screening method for AR, we developed a deep learning-based artificial intelligence algorithm for the diagnosis of significant AR using electrocardiography (ECG). METHODS: Our dataset comprised 29,859 paired data of ECG and echocardiography, including 412 AR cases, from January 2015 to December 2019. This dataset was divided into training, validation, and test datasets. We developed a multi-input neural network model, which comprised a two-dimensional convolutional neural network (2D-CNN) using raw ECG data and a fully connected deep neural network (FC-DNN) using ECG features, and compared its performance with the performances of a 2D-CNN model and other machine learning models. In addition, we used gradient-weighted class activation mapping (Grad-CAM) to identify which parts of ECG waveforms had the most effect on algorithm decision making. RESULTS: The area under the receiver operating characteristic curve of the multi-input model (0.802; 95% CI, 0.762-0.837) was significantly greater than that of the 2D-CNN model alone (0.734; 95% CI, 0.679-0.783; p<0.001) and those of other machine learning models. Grad-CAM demonstrated that the multi-input model tended to focus on the QRS complex in leads I and aVL when detecting AR. CONCLUSIONS: The multi-input deep learning model using 12-lead ECG data could detect significant AR with modest predictive value.

Treatment Discontinuation Prediction in Patients With Diabetes Using a Ranking Model: Machine Learning Model Development.

BACKGROUND: Treatment discontinuation (TD) is one of the major prognostic issues in diabetes care, and several models have been proposed to predict a missed appointment that may lead to TD in patients with diabetes by using binary classification models for the early detection of TD and for providing intervention support for patients. However, as binary classification models output the probability of a missed appointment occurring within a predetermined period, they are limited in their ability to estimate the magnitude of TD risk in patients with inconsistent intervals between appointments, making it difficult to prioritize patients for whom intervention support should be provided. OBJECTIVE: This study aimed to develop a machine-learned prediction model that can output a TD risk score defined by the length of time until TD and prioritize patients for intervention according to their TD risk. METHODS: This model included patients with diagnostic codes indicative of diabetes at the University of Tokyo Hospital between September 3, 2012, and May 17, 2014. The model was internally validated with patients from the same hospital from May 18, 2014, to January 29, 2016. The data used in this study included 7551 patients who visited the hospital after January 1, 2004, and had diagnostic codes indicative of diabetes. In particular, data that were recorded in the electronic medical records between September 3, 2012, and January 29, 2016, were used. The main outcome was the TD of a patient, which was defined as missing a scheduled clinical appointment and having no hospital visits within 3 times the average number of days between the visits of the patient and within 60 days. The TD risk score was calculated by using the parameters derived from the machine-learned ranking model. The prediction capacity was evaluated by using test data with the C-index for the performance of ranking patients, area under the receiver operating characteristic curve, and area under the precision-recall curve for discrimination, in addition to a calibration plot. RESULTS: The means (95% confidence limits) of the C-index, area under the receiver operating characteristic curve, and area under the precision-recall curve for the TD risk score were 0.749 (0.655, 0.823), 0.758 (0.649, 0.857), and 0.713 (0.554, 0.841), respectively. The observed and predicted probabilities were correlated with the calibration plots. CONCLUSIONS: A TD risk score was developed for patients with diabetes by combining a machine-learned method with electronic medical records. The score calculation can be integrated into medical records to identify patients at high risk of TD, which would be useful in supporting diabetes care and preventing TD.

The Effectiveness of a Deep Learning Model to Detect Left Ventricular Systolic Dysfunction from Electrocardiograms.

Deep learning models can be applied to electrocardiograms (ECGs) to detect left ventricular (LV) dysfunction. We hypothesized that applying a deep learning model may improve the diagnostic accuracy of cardiologists in predicting LV dysfunction from ECGs. We acquired 37,103 paired ECG and echocardiography data records of patients who underwent echocardiography between January 2015 and December 2019. We trained a convolutional neural network to identify the data records of patients with LV dysfunction (ejection fraction < 40%) using a dataset of 23,801 ECGs. When tested on an independent set of 7,196 ECGs, we found the area under the receiver operating characteristic curve was 0.945 (95% confidence interval: 0.936-0.954). When 7 cardiologists interpreted 50 randomly selected ECGs from the test dataset of 7,196 ECGs, their accuracy for predicting LV dysfunction was 78.0% ± 6.0%. By referring to the model's output, the cardiologist accuracy improved to 88.0% ± 3.7%, which indicates that model support significantly improved the cardiologist diagnostic accuracy (P = 0.02). A sensitivity map demonstrated that the model focused on the QRS complex when detecting LV dysfunction on ECGs. We developed a deep learning model that can detect LV dysfunction on ECGs with high accuracy. Furthermore, we demonstrated that support from a deep learning model can help cardiologists to identify LV dysfunction on ECGs.

Machine learning-based prediction of in-hospital mortality using admission laboratory data: A retrospective, single-site study using electronic health record data.

Risk assessment of in-hospital mortality of patients at the time of hospitalization is necessary for determining the scale of required medical resources for the patient depending on the patient's severity. Because recent machine learning application in the clinical area has been shown to enhance prediction ability, applying this technique to this issue can lead to an accurate prediction model for in-hospital mortality prediction. In this study, we aimed to generate an accurate prediction model of in-hospital mortality using machine learning techniques. Patients 18 years of age or older admitted to the University of Tokyo Hospital between January 1, 2009 and December 26, 2017 were used in this study. The data were divided into a training/validation data set (n = 119,160) and a test data set (n = 33,970) according to the time of admission. The prediction target of the model was the in-hospital mortality within 14 days. To generate the prediction model, 25 variables (age, sex, 21 laboratory test items, length of stay, and mortality) were used to predict in-hospital mortality. Logistic regression, random forests, multilayer perceptron, and gradient boost decision trees were performed to generate the prediction models. To evaluate the prediction capability of the model, the model was tested using a test data set. Mean probabilities obtained from trained models with five-fold cross-validation were used to calculate the area under the receiver operating characteristic (AUROC) curve. In a test stage using the test data set, prediction models of in-hospital mortality within 14 days showed AUROC values of 0.936, 0.942, 0.942, and 0.938 for logistic regression, random forests, multilayer perceptron, and gradient boosting decision trees, respectively. Machine learning-based prediction of short-term in-hospital mortality using admission laboratory data showed outstanding prediction capability and, therefore, has the potential to be useful for the risk assessment of patients at the time of hospitalization.

Anti-senescent drug screening by deep learning-based morphology senescence scoring.

Advances in deep learning technology have enabled complex task solutions. The accuracy of image classification tasks has improved owing to the establishment of convolutional neural networks (CNN). Cellular senescence is a hallmark of ageing and is important for the pathogenesis of ageing-related diseases. Furthermore, it is a potential therapeutic target. Specific molecular markers are used to identify senescent cells. Moreover senescent cells show unique morphology, which can be identified. We develop a successful morphology-based CNN system to identify senescent cells and a quantitative scoring system to evaluate the state of endothelial cells by senescence probability output from pre-trained CNN optimised for the classification of cellular senescence, Deep Learning-Based Senescence Scoring System by Morphology (Deep-SeSMo). Deep-SeSMo correctly evaluates the effects of well-known anti-senescent reagents. We screen for drugs that control cellular senescence using a kinase inhibitor library by Deep-SeSMo-based drug screening and identify four anti-senescent drugs. RNA sequence analysis reveals that these compounds commonly suppress senescent phenotypes through inhibition of the inflammatory response pathway. Thus, morphology-based CNN system can be a powerful tool for anti-senescent drug screening.

Outcome prediction of out-of-hospital cardiac arrest with presumed cardiac aetiology using an advanced machine learning technique.

BACKGROUND: Outcome prediction for patients with out-of-hospital cardiac arrest (OHCA) has the possibility to detect patients who could have been potentially saved. Advanced machine learning techniques have recently been developed and employed for clinical studies. In this study, we aimed to establish a prognostication model for OHCA with presumed cardiac aetiology using an advanced machine learning technique. METHODS AND RESULTS: Cohort data from a prospective multi-centre cohort study for OHCA patients transported by an ambulance in the Kanto area of Japan between January 2012 and March 2013 (SOS-KANTO 2012 study) were analysed in this study. Of 16,452 patients, data for OHCA patients aged ≥18 years with presumed cardiac aetiology were retrieved, and were divided into two groups (training set: n = 5718, between January 1, 2012 and December 12, 2012; test set: n = 1608, between January 1, 2013 and March 31, 2013). Of 421 variables observed during prehospital and emergency department settings, 35 prehospital variables, or 35 prehospital and 18 in-hospital variables, were used for outcome prediction of 1-year survival using a random forest method. In validation using the test set, prognostication models trained with 35 variables, or 53 variables for 1-year survival showed area under the receiver operating characteristics curve (AUC) values of 0.943 (95% CI [0.930, 0.955]) and 0.958 (95% CI [0.948, 0.969]), respectively. CONCLUSIONS: The advanced machine learning technique showed favourable prediction capability for 1-year survival of OHCA with presumed cardiac aetiology. These models can be useful for detecting patients who could have been potentially saved.

Selective elimination of undifferentiated human pluripotent stem cells using pluripotent state-specific immunogenic antigen Glypican-3.

Immunogenicity of immature pluripotent stem cells is a topic of intense debate. Immunogenic antigens, which are specific in pluripotent states, have not been described previously. In this study, we identified glypican-3 (GPC3), a known carcinoembryonic antigen, as a pluripotent state-specific immunogenic antigen. Additionally, we validated the applicability of human leukocyte antigen (HLA)-class I-restricted GPC3-reactive cytotoxic T lymphocytes (CTLs) in the removal of undifferentiated pluripotent stem cells (PSCs) from human induced pluripotent stem cell (hiPSC)-derivatives. HiPSCs uniquely express GPC3 in pluripotent states and were rejected by GPC3-reactive CTLs, which were sensitized with HLA-class I-restricted GPC3 peptides. Furthermore, GPC3-reactive CTLs selectively removed undifferentiated PSCs from hiPSC-derivatives in vitro and inhibited tumor formation in vivo. Our results demonstrate that GPC3 works as a pluripotent state-specific immunogenic antigen in hiPSCs and is applicable to regenerative medicine as a method of removing undifferentiated PSCs, which are the main cause of tumor formation.

Development of a transplant injection device for optimal distribution and retention of human induced pluripotent stem cell‒derived cardiomyocytes.

BACKGROUND: Induced pluripotent stem cell (iPSC)‒based regenerative therapy is a promising strategy for cardiovascular disease treatment; however, the method is limited by the myocardial retention of grafted iPSCs. Thus, an injection protocol that efficiently introduces and retains human iPSC-derived cardiomyocytes (hiPSC-CMs) within the myocardium is urgently needed. The objective of the present study was to develop a method to improve the retention of hiPSCs in the myocardium for cardiac therapy. METHODS: We efficiently produced hiPSC-CM spheroids in 3-dimensional (3D) culture using microwell plates, and developed an injection device for optimal 3D distribution of the spheroids in the myocardial layer. Device biocompatibility was assessed with purified hiPSC-CM spheroids. Device effectiveness was evaluated in 10- to 15-month-old farm pigs (n = 15) and 5- to 24-month-old micro-minipigs (n = 20). The pigs were euthanized after injection, and tissues were harvested for retention and histologic analysis. RESULTS: We demonstrated an injection device for direct intramyocardial transplantation of hiPSC-CM spheroids from large-scale culture. The device had no detrimental effects on cell viability, spheroid shape, or size. Direct epicardial injection of spheroids mixed with gelatin hydrogel into beating porcine hearts using this device resulted in better distribution and retention of transplanted spheroids in a layer within the myocardium than did conventional needle injection procedures. CONCLUSIONS: The combination of the newly developed transplant device and spheroid formation promotes the retention of transplanted CMs. These findings support the clinical application of hiPSC-CM spheroid‒based cardiac regenerative therapy in patients with heart failure.