Predicting intubation for intensive care units patients: A deep learning approach to improve patient management.

OBJECTIVE: For patients in the Intensive Care Unit (ICU), the timing of intubation has a significant association with patients' outcomes. However, accurate prediction of the timing of intubation remains an unsolved challenge due to the noisy, sparse, heterogeneous, and unbalanced nature of ICU data. In this study, our objective is to develop a workflow for pre-processing ICU data and to develop a customized deep learning model to predict the need for intubation. METHODS: To improve the prediction accuracy, we transform the intubation prediction task into a time series classification task. We carefully design a sequence of data pre-processing steps to handle the multimodal noisy data. Firstly, we discretize the sequential data and address missing data using interpolation. Next, we employ a sampling strategy to address data imbalance and standardize the data to facilitate faster model convergence. Furthermore, we employ the feature selection technique and propose an ensemble model to combine features learned by different deep learning models. RESULTS: The performance is evaluated on Medical Information Mart for Intensive Care (MIMIC)-III, an ICU dataset. Our proposed Deep Feature Fusion method achieves an area under the curve (AUC) of the receiver operating curve (ROC) of 0.8953, surpassing the performance of other deep learning and traditional machine learning models. CONCLUSION: Our proposed Deep Feature Fusion method proves to be a viable approach for predicting intubation and outperforms other deep learning and classical machine learning models. The study confirms that high-frequency time-varying indicators, particularly Mean Blood Pressure (MeanBP) and peripheral oxygen saturation (SpO2), are significant risk factors for predicting intubation.

A machine-learning exploration of the exposome from preconception in early childhood atopic eczema, rhinitis and wheeze development.

BACKGROUND: Most previous research on the environmental epidemiology of childhood atopic eczema, rhinitis and wheeze is limited in the scope of risk factors studied. Our study adopted a machine learning approach to explore the role of the exposome starting already in the preconception phase. METHODS: We performed a combined analysis of two multi-ethnic Asian birth cohorts, the Growing Up in Singapore Towards healthy Outcomes (GUSTO) and the Singapore PREconception Study of long Term maternal and child Outcomes (S-PRESTO) cohorts. Interviewer-administered questionnaires were used to collect information on demography, lifestyle and childhood atopic eczema, rhinitis and wheeze development. Data training was performed using XGBoost, genetic algorithm and logistic regression models, and the top variables with the highest importance were identified. Additive explanation values were identified and inputted into a final multiple logistic regression model. Generalised structural equation modelling with maternal and child blood micronutrients, metabolites and cytokines was performed to explain possible mechanisms. RESULTS: The final study population included 1151 mother-child pairs. Our findings suggest that these childhood diseases are likely programmed in utero by the preconception and pregnancy exposomes through inflammatory pathways. We identified preconception alcohol consumption and maternal depressive symptoms during pregnancy as key modifiable maternal environmental exposures that increased eczema and rhinitis risk. Our mechanistic model suggested that higher maternal blood neopterin and child blood dimethylglycine protected against early childhood wheeze. After birth, early infection was a key driver of atopic eczema and rhinitis development. CONCLUSION: Preconception and antenatal exposomes can programme atopic eczema, rhinitis and wheeze development in utero. Reducing maternal alcohol consumption during preconception and supporting maternal mental health during pregnancy may prevent atopic eczema and rhinitis by promoting an optimal antenatal environment. Our findings suggest a need to include preconception environmental exposures in future research to counter the earliest precursors of disease development in children.

Association of body temperature and mortality in critically ill patients: an observational study using two large databases.

BACKGROUND: Body temperature (BT) is routinely measured and can be controlled in critical care settings. BT can impact patient outcome, but the relationship between BT and mortality has not been well-established. METHODS: A retrospective cohort study was conducted based on the MIMIC-IV (N = 43,537) and eICU (N = 75,184) datasets. The primary outcome and exposure variables were hospital mortality and first 48-h median BT, respectively. Generalized additive models were used to model the associations between exposures and outcomes, while adjusting for patient age, sex, APS-III, SOFA, and Charlson comorbidity scores, temperature gap, as well as ventilation, vasopressor, steroids, and dialysis usage. We conducted subgroup analysis according to ICU setting, diagnoses, and demographics. RESULTS: Optimal BT was 37 °C for the general ICU and subgroup populations. A 10% increase in the proportion of time that BT was within the 36-38 °C range was associated with reduced hospital mortality risk in both MIMIC-IV (OR 0.91; 95% CI 0.90-0.93) and eICU (OR 0.86; 95% CI 0.85-0.87). On the other hand, a 10% increase in the proportion of time when BT < 36 °C was associated with increased mortality risk in both MIMIC-IV (OR 1.08; 95% CI 1.06-1.10) and eICU (OR 1.18; 95% CI 1.16-1.19). Similarly, a 10% increase in the proportion of time when BT > 38 °C was associated with increased mortality risk in both MIMIC-IV (OR 1.09; 95% CI 1.07-1.12) and eICU (OR 1.09; 95% CI 1.08-1.11). All patient subgroups tested consistently showed an optimal temperature within the 36-38 °C range. CONCLUSIONS: A BT of 37 °C is associated with the lowest mortality risk among ICU patients. Further studies to explore the causal relationship between the optimal BT and mortality should be conducted and may help with establishing guidelines for active BT management in critical care settings.

The SingHealth Perioperative and Anesthesia Subject Area Registry (PASAR), a large-scale perioperative data mart and registry.

BACKGROUND: To enhance perioperative outcomes, a perioperative registry that integrates high-quality real-world data throughout the perioperative period is essential. Singapore General Hospital established the Perioperative and Anesthesia Subject Area Registry (PASAR) to unify data from the preoperative, intraoperative, and postoperative stages. This study presents the methodology employed to create this database. METHODS: Since 2016, data from surgical patients have been collected from the hospital electronic medical record systems, de-identified, and stored securely in compliance with privacy and data protection laws. As a representative sample, data from initiation in 2016 to December 2022 were collected. RESULTS: As of December 2022, PASAR data comprise 26 tables, encompassing 153,312 patient admissions and 168,977 operation sessions. For this period, the median age of the patients was 60.0 years, sex distribution was balanced, and the majority were Chinese. Hypertension and cardiovascular comorbidities were also prevalent. Information including operation type and time, intensive care unit (ICU) length of stay, and 30-day and 1-year mortality rates were collected. Emergency surgeries resulted in longer ICU stays, but shorter operation times than elective surgeries. CONCLUSIONS: The PASAR provides a comprehensive and automated approach to gathering high-quality perioperative patient data.

A fully convolutional neural network for comprehensive compartmentalization of abdominal adipose tissue compartments in MRI.

BACKGROUND: Existing literature has highlighted structural, physiological, and pathological disparities among abdominal adipose tissue (AAT) sub-depots. Accurate separation and quantification of these sub-depots are crucial for advancing our understanding of obesity and its comorbidities. However, the absence of clear boundaries between the sub-depots in medical imaging data has challenged their separation, particularly for internal adipose tissue (IAT) sub-depots. To date, the quantification of AAT sub-depots remains challenging, marked by a time-consuming, costly, and complex process. PURPOSE: To implement and evaluate a convolutional neural network to enable granular assessment of AAT by compartmentalization of subcutaneous adipose tissue (SAT) into superficial subcutaneous (SSAT) and deep subcutaneous (DSAT) adipose tissue, and IAT into intraperitoneal (IPAT), retroperitoneal (RPAT), and paraspinal (PSAT) adipose tissue. MATERIAL AND METHODS: MRI datasets were retrospectively collected from Singapore Preconception Study for Long-Term Maternal and Child Outcomes (S-PRESTO: 389 women aged 31.4 ± 3.9 years) and Singapore Adult Metabolism Study (SAMS: 50 men aged 28.7 ± 5.7 years). For all datasets, ground truth segmentation masks were created through manual segmentation. A Res-Net based 3D-UNet was trained and evaluated via 5-fold cross-validation on S-PRESTO data (N = 300). The model's final performance was assessed on a hold-out (N = 89) and an external test set (N = 50, SAMS). RESULTS: The proposed method enabled reliable segmentation of individual AAT sub-depots in 3D MRI volumes with high mean Dice similarity scores of 98.3%, 97.2%, 96.5%, 96.3%, and 95.9% for SSAT, DSAT, IPAT, RPAT, and PSAT respectively. CONCLUSION: Convolutional neural networks can accurately sub-divide abdominal SAT into SSAT and DSAT, and abdominal IAT into IPAT, RPAT, and PSAT with high accuracy. The presented method has the potential to significantly contribute to advancements in the field of obesity imaging and precision medicine.

Data-Efficient Computational Pathology Platform for Faster and Cheaper Breast Cancer Subtype Identifications: Development of a Deep Learning Model.

BACKGROUND: Breast cancer subtyping is a crucial step in determining therapeutic options, but the molecular examination based on immunohistochemical staining is expensive and time-consuming. Deep learning opens up the possibility to predict the subtypes based on the morphological information from hematoxylin and eosin staining, a much cheaper and faster alternative. However, training the predictive model conventionally requires a large number of histology images, which is challenging to collect by a single institute. OBJECTIVE: We aimed to develop a data-efficient computational pathology platform, 3DHistoNet, which is capable of learning from z-stacked histology images to accurately predict breast cancer subtypes with a small sample size. METHODS: We retrospectively examined 401 cases of patients with primary breast carcinoma diagnosed between 2018 and 2020 at the Department of Pathology, National Cancer Center, South Korea. Pathology slides of the patients with breast carcinoma were prepared according to the standard protocols. Age, gender, histologic grade, hormone receptor (estrogen receptor [ER], progesterone receptor [PR], and androgen receptor [AR]) status, erb-B2 receptor tyrosine kinase 2 (HER2) status, and Ki-67 index were evaluated by reviewing medical charts and pathological records. RESULTS: The area under the receiver operating characteristic curve and decision curve were analyzed to evaluate the performance of our 3DHistoNet platform for predicting the ER, PR, AR, HER2, and Ki67 subtype biomarkers with 5-fold cross-validation. We demonstrated that 3DHistoNet can predict all clinically important biomarkers (ER, PR, AR, HER2, and Ki67) with performance exceeding the conventional multiple instance learning models by a considerable margin (area under the receiver operating characteristic curve: 0.75-0.91 vs 0.67-0.8). We further showed that our z-stack histology scanning method can make up for insufficient training data sets without any additional cost incurred. Finally, 3DHistoNet offered an additional capability to generate attention maps that reveal correlations between Ki67 and histomorphological features, which renders the hematoxylin and eosin image in higher fidelity to the pathologist. CONCLUSIONS: Our stand-alone, data-efficient pathology platform that can both generate z-stacked images and predict key biomarkers is an appealing tool for breast cancer diagnosis. Its development would encourage morphology-based diagnosis, which is faster, cheaper, and less error-prone compared to the protein quantification method based on immunohistochemical staining.

Illness severity assessment of older adults in critical illness using machine learning (ELDER-ICU): an international multicentre study with subgroup bias evaluation.

BACKGROUND: Comorbidity, frailty, and decreased cognitive function lead to a higher risk of death in elderly patients (more than 65 years of age) during acute medical events. Early and accurate illness severity assessment can support appropriate decision making for clinicians caring for these patients. We aimed to develop ELDER-ICU, a machine learning model to assess the illness severity of older adults admitted to the intensive care unit (ICU) with cohort-specific calibration and evaluation for potential model bias. METHODS: In this retrospective, international multicentre study, the ELDER-ICU model was developed using data from 14 US hospitals, and validated in 171 hospitals from the USA and Netherlands. Data were extracted from the Medical Information Mart for Intensive Care database, electronic ICU Collaborative Research Database, and Amsterdam University Medical Centers Database. We used six categories of data as predictors, including demographics and comorbidities, physical frailty, laboratory tests, vital signs, treatments, and urine output. Patient data from the first day of ICU stay were used to predict in-hospital mortality. We used the eXtreme Gradient Boosting algorithm (XGBoost) to develop models and the SHapley Additive exPlanations method to explain model prediction. The trained model was calibrated before internal, external, and temporal validation. The final XGBoost model was compared against three other machine learning algorithms and five clinical scores. We performed subgroup analysis based on age, sex, and race. We assessed the discrimination and calibration of models using the area under receiver operating characteristic (AUROC) and standardised mortality ratio (SMR) with 95% CIs. FINDINGS: Using the development dataset (n=50 366) and predictive model building process, the XGBoost algorithm performed the best in all types of validations compared with other machine learning algorithms and clinical scores (internal validation with 5037 patients from 14 US hospitals, AUROC=0·866 [95% CI 0·851-0·880]; external validation in the US population with 20 541 patients from 169 hospitals, AUROC=0·838 [0·829-0·847]; external validation in European population with 2411 patients from one hospital, AUROC=0·833 [0·812-0·853]; temporal validation with 4311 patients from one hospital, AUROC=0·884 [0·869-0·897]). In the external validation set (US population), the median AUROCs of bias evaluations covering eight subgroups were above 0·81, and the overall SMR was 0·99 (0·96-1·03). The top ten risk predictors were the minimum Glasgow Coma Scale score, total urine output, average respiratory rate, mechanical ventilation use, best state of activity, Charlson Comorbidity Index score, geriatric nutritional risk index, code status, age, and maximum blood urea nitrogen. A simplified model containing only the top 20 features (ELDER-ICU-20) had similar predictive performance to the full model. INTERPRETATION: The ELDER-ICU model reliably predicts the risk of in-hospital mortality using routinely collected clinical features. The predictions could inform clinicians about patients who are at elevated risk of deterioration. Prospective validation of this model in clinical practice and a process for continuous performance monitoring and model recalibration are needed. FUNDING: National Institutes of Health, National Natural Science Foundation of China, National Special Health Science Program, Health Science and Technology Plan of Zhejiang Province, Fundamental Research Funds for the Central Universities, Drug Clinical Evaluate Research of Chinese Pharmaceutical Association, and National Key R&D Program of China.

Artificial intelligence and machine learning in prehospital emergency care: A scoping review.

Our scoping review provides a comprehensive analysis of the landscape of artificial intelligence (AI) applications in prehospital emergency care (PEC). It contributes to the field by highlighting the most studied AI applications and identifying the most common methodological approaches across 106 included studies. The findings indicate a promising future for AI in PEC, with many unique use cases, such as prognostication, demand prediction, resource optimization, and the Internet of Things continuous monitoring systems. Comparisons with other approaches showed AI outperforming clinicians and non-AI algorithms in most cases. However, most studies were internally validated and retrospective, highlighting the need for rigorous prospective validation of AI applications before implementation in clinical settings. We identified knowledge and methodological gaps using an evidence map, offering a roadmap for future investigators. We also discussed the significance of explainable AI for establishing trust in AI systems among clinicians and facilitating real-world validation of AI models.

Introducing value-based healthcare perspectives into hospital performance assessment: A scoping review.

OBJECTIVE: Value-based healthcare (VBHC) puts patient outcomes at the center of the healthcare process while optimizing the use of hospital resources across multiple stakeholders. This scoping review was conducted to summarize how VBHC had been represented in theory and in practice, how it had been applied to assess hospital performance, and how well it had been ultimately implemented. METHODS: For this review, we followed the PRISMA-ScR protocol and searched five major online databases for articles published between January 2006 and July 2022. We included original articles that used the concept of VBHC to conduct performance assessments of healthcare organizations. We extracted and analyzed key concepts and information on the dimensions of VBHC, specific strategies and methods for using VBHC in performance assessment, and the effectiveness of the assessment. RESULTS: We identified 48 eligible studies from 7866 articles. Nineteen nonempirical studies focused on the development of a VBHC performance assessment indicator system, and 29 empirical studies reported on the ways and points of introducing VBHC into performance assessment and its effectiveness. Ultimately, we summarized the key dimensions, processes, and effects of performance assessment after introducing VBHC. CONCLUSION: Current healthcare performance assessment has begun to focus on implementing VBHC as an integrated strategy, and future work should further clarify the reliability of metrics and their association with evaluation outcomes and consider the effective integration of clinical outcomes and patient-reported outcomes.

Automatic Calcification Morphology and Distribution Classification for Breast Mammograms With Multi-Task Graph Convolutional Neural Network.

The morphology and distribution of microcalcifications are the most important descriptors for radiologists to diagnose breast cancer based on mammograms. However, it is very challenging and time-consuming for radiologists to characterize these descriptors manually, and there also lacks of effective and automatic solutions for this problem. We observed that the distribution and morphology descriptors are determined by the radiologists based on the spatial and visual relationships among calcifications. Thus, we hypothesize that this information can be effectively modelled by learning a relationship-aware representation using graph convolutional networks (GCNs). In this study, we propose a multi-task deep GCN method for automatic characterization of both the morphology and distribution of microcalcifications in mammograms. Our proposed method transforms morphology and distribution characterization into node and graph classification problem and learns the representations concurrently. We trained and validated the proposed method in an in-house dataset and public DDSM dataset with 195 and 583 cases,respectively. The proposed method reaches good and stable results with distribution AUC at 0.812 ± 0.043 and 0.873 ± 0.019, morphology AUC at 0.663 ± 0.016 and 0.700 ± 0.044 for both in-house and public datasets. In both datasets, our proposed method demonstrates statistically significant improvements compared to the baseline models. The performance improvements brought by our proposed multi-task mechanism can be attributed to the association between the distribution and morphology of calcifications in mammograms, which is interpretable using graphical visualizations and consistent with the definitions of descriptors in the standard BI-RADS guideline. In short, we explore, for the first time, the application of GCNs in microcalcification characterization that suggests the potential of using graph learning for more robust understanding of medical images.

Differential Diagnosis of DCIS and Fibroadenoma Based on Ultrasound Images: a Difference-Based Self-Supervised Approach.

Differentiation of ductal carcinoma in situ (DCIS, a precancerous lesion of the breast) from fibroadenoma (FA) using ultrasonography is significant for the early prevention of malignant breast tumors. Radiomics-based artificial intelligence (AI) can provide additional diagnostic information but usually requires extensive labeling efforts by clinicians with specialized knowledge. This study aims to investigate the feasibility of differentially diagnosing DCIS and FA using ultrasound radiomics-based AI techniques and further explore a novel approach that can reduce labeling efforts without sacrificing diagnostic performance. We included 461 DCIS and 651 FA patients, of whom 139 DCIS and 181 FA patients constituted a prospective test cohort. First, various feature engineering-based machine learning (FEML) and deep learning (DL) approaches were developed. Then, we designed a difference-based self-supervised (DSS) learning approach that only required FA samples to participate in training. The DSS approach consists of three steps: (1) pretraining a Bootstrap Your Own Latent (BYOL) model using FA images, (2) reconstructing images using the encoder and decoder of the pretrained model, and (3) distinguishing DCIS from FA based on the differences between the original and reconstructed images. The experimental results showed that the trained FEML and DL models achieved the highest AUC of 0.7935 (95% confidence interval, 0.7900-0.7969) on the prospective test cohort, indicating that the developed models are effective for assisting in differentiating DCIS from FA based on ultrasound images. Furthermore, the DSS model achieved an AUC of 0.8172 (95% confidence interval, 0.8124-0.8219), indicating that our model outperforms the conventional radiomics-based AI models and is more competitive.

Building a Machine Learning-based Ambulance Dispatch Triage Model for Emergency Medical Services.

Background: In charge of dispatching the ambulances, Emergency Medical Services (EMS) call center specialists often have difficulty deciding the acuity of a case given the information they can gather within a limited time. Although there are protocols to guide their decision-making, observed performance can still lack sensitivity and specificity. Machine learning models have been known to capture complex relationships that are subtle, and well-trained data models can yield accurate predictions in a split of a second. Methods: In this study, we proposed a proof-of-concept approach to construct a machine learning model to better predict the acuity of emergency cases. We used more than 360,000 structured emergency call center records of cases received by the national emergency call center in Singapore from 2018 to 2020. Features were created using call records, and multiple machine learning models were trained. Results: A Random Forest model achieved the best performance, reducing the over-triage rate by an absolute margin of 15% compared to the call center specialists while maintaining a similar level of under-triage rate. Conclusions: The model has the potential to be deployed as a decision support tool for dispatchers alongside current protocols to optimize ambulance dispatch triage and the utilization of emergency ambulance resources.

The application of artificial intelligence in the management of sepsis.

Sepsis is a complex and heterogeneous syndrome that remains a serious challenge to healthcare worldwide. Patients afflicted by severe sepsis or septic shock are customarily placed under intensive care unit (ICU) supervision, where a multitude of apparatus is poised to produce high-granularity data. This reservoir of high-quality data forms the cornerstone for the integration of AI into clinical practice. However, existing reviews currently lack the inclusion of the latest advancements. This review examines the evolving integration of artificial intelligence (AI) in sepsis management. Applications of artificial intelligence include early detection, subtyping analysis, precise treatment and prognosis assessment. AI-driven early warning systems provide enhanced recognition and intervention capabilities, while profiling analyzes elucidate distinct sepsis manifestations for targeted therapy. Precision medicine harnesses the potential of artificial intelligence for pathogen identification, antibiotic selection, and fluid optimization. In conclusion, the seamless amalgamation of artificial intelligence into the domain of sepsis management heralds a transformative shift, ushering in novel prospects to elevate diagnostic precision, therapeutic efficacy, and prognostic acumen. As AI technologies develop, their impact on shaping the future of sepsis care warrants ongoing research and thoughtful implementation.

FCSN: Global Context Aware Segmentation by Learning the Fourier Coefficients of Objects in Medical Images.

The encoder-decoder model is a commonly used Deep Neural Network (DNN) model for medical image segmentation. Conventional encoder-decoder models make pixel-wise predictions focusing heavily on local patterns around the pixel. This makes it challenging to give segmentation that preserves the object's shape and topology, which often requires an understanding of the global context. In this work, we propose a Fourier Coefficient Segmentation Network (FCSN)-a novel global context-aware DNN model that segments an object by learning the complex Fourier coefficients of the object's masks. The Fourier coefficients are calculated by integrating over the whole contour. Therefore, for our model to make a precise estimation of the coefficients, the model is motivated to incorporate the global context of the object, leading to a more accurate segmentation of the object's shape. This global context awareness also makes our model robust to unseen local perturbations during inference, such as additive noise or motion blur that are prevalent in medical images. We compare FCSN with other state-of-the-art global context-aware models (UNet++, DeepLabV3+, UNETR) on 5 medical image segmentation tasks, of which 3 are camera imaging datasets (ISIC_2018, RIM_CUP, RIM_DISC) and 2 are medical imaging datasets (PROSTATE, FETAL). When FCSN is compared with UNETR, FCSN attains significantly lower Hausdorff scores with 19.14 (6%), 17.42 (6%), 9.16 (14%), 11.18 (22%), and 5.98 (6%) for ISIC_2018, RIM_CUP, RIM_DISC, PROSTATE, and FETAL tasks respectively. Moreover, FCSN is lightweight by discarding the decoder module, which incurs significant computational overhead. FCSN only requires 29.7 M parameters which are 75.6 M and 9.9 M fewer parameters than UNETR and DeepLabV3+, respectively. FCSN attains inference and training speeds of 1.6 ms/img and 6.3 ms/img, which is 8× and 3× faster than UNet and UNETR. The code for FCSN is made publicly available at https://github.com/nus-mornin-lab/FCSN.

UFRC: A Unified Framework for Reliable COVID-19 Detection on Crowdsourced Cough Audio.

We suggested a unified system with core components of data augmentation, ImageNet-pretrained ResNet-50, cost-sensitive loss, deep ensemble learning, and uncertainty estimation to quickly and consistently detect COVID-19 using acoustic evidence. To increase the model's capacity to identify a minority class, data augmentation and cost-sensitive loss are incorporated (infected samples). In the COVID-19 detection challenge, ImageNet-pretrained ResNet-50 has been found to be effective. The unified framework also integrates deep ensemble learning and uncertainty estimation to integrate predictions from various base classifiers for generalisation and reliability. We ran a series of tests using the DiCOVA2021 challenge dataset to assess the efficacy of our proposed method, and the results show that our method has an AUC-ROC of 85.43 percent, making it a promising method for COVID-19 detection. The unified framework also demonstrates that audio may be used to quickly diagnose different respiratory disorders.

Machine Learning-Derived Prenatal Predictive Risk Model to Guide Intervention and Prevent the Progression of Gestational Diabetes Mellitus to Type 2 Diabetes: Prediction Model Development Study.

BACKGROUND: The increasing prevalence of gestational diabetes mellitus (GDM) is concerning as women with GDM are at high risk of type 2 diabetes (T2D) later in life. The magnitude of this risk highlights the importance of early intervention to prevent the progression of GDM to T2D. Rates of postpartum screening are suboptimal, often as low as 13% in Asian countries. The lack of preventive care through structured postpartum screening in several health care systems and low public awareness are key barriers to postpartum diabetes screening. OBJECTIVE: In this study, we developed a machine learning model for early prediction of postpartum T2D following routine antenatal GDM screening. The early prediction of postpartum T2D during prenatal care would enable the implementation of effective strategies for diabetes prevention interventions. To our best knowledge, this is the first study that uses machine learning for postpartum T2D risk assessment in antenatal populations of Asian origin. METHODS: Prospective multiethnic data (Chinese, Malay, and Indian ethnicities) from 561 pregnancies in Singapore's most deeply phenotyped mother-offspring cohort study-Growing Up in Singapore Towards healthy Outcomes-were used for predictive modeling. The feature variables included were demographics, medical or obstetric history, physical measures, lifestyle information, and GDM diagnosis. Shapley values were combined with CatBoost tree ensembles to perform feature selection. Our game theoretical approach for predictive analytics enables population subtyping and pattern discovery for data-driven precision care. The predictive models were trained using 4 machine learning algorithms: logistic regression, support vector machine, CatBoost gradient boosting, and artificial neural network. We used 5-fold stratified cross-validation to preserve the same proportion of T2D cases in each fold. Grid search pipelines were built to evaluate the best performing hyperparameters. RESULTS: A high performance prediction model for postpartum T2D comprising of 2 midgestation features-midpregnancy BMI after gestational weight gain and diagnosis of GDM-was developed (BMI_GDM CatBoost model: AUC=0.86, 95% CI 0.72-0.99). Prepregnancy BMI alone was inadequate in predicting postpartum T2D risk (ppBMI CatBoost model: AUC=0.62, 95% CI 0.39-0.86). A 2-hour postprandial glucose test (BMI_2hour CatBoost model: AUC=0.86, 95% CI 0.76-0.96) showed a stronger postpartum T2D risk prediction effect compared to fasting glucose test (BMI_Fasting CatBoost model: AUC=0.76, 95% CI 0.61-0.91). The BMI_GDM model was also robust when using a modified 2-point International Association of the Diabetes and Pregnancy Study Groups (IADPSG) 2018 criteria for GDM diagnosis (BMI_GDM2 CatBoost model: AUC=0.84, 95% CI 0.72-0.97). Total gestational weight gain was inversely associated with postpartum T2D outcome, independent of prepregnancy BMI and diagnosis of GDM (P=.02; OR 0.88, 95% CI 0.79-0.98). CONCLUSIONS: Midgestation weight gain effects, combined with the metabolic derangements underlying GDM during pregnancy, signal future T2D risk in Singaporean women. Further studies will be required to examine the influence of metabolic adaptations in pregnancy on postpartum maternal metabolic health outcomes. The state-of-the-art machine learning model can be leveraged as a rapid risk stratification tool during prenatal care. TRIAL REGISTRATION: ClinicalTrials.gov NCT01174875; https://clinicaltrials.gov/ct2/show/NCT01174875.

End-to-End Calcification Distribution Pattern Recognition for Mammograms: An Interpretable Approach with GNN.

Purpose: We aimed to develop a novel interpretable artificial intelligence (AI) model algorithm focusing on automatic detection and classification of various patterns of calcification distribution in mammographic images using a unique graph convolution approach. Materials and methods: Images from 292 patients, which showed calcifications according to the mammographic reports and diagnosed breast cancers, were collected. The calcification distributions were classified as diffuse, segmental, regional, grouped, or linear. Excluded were mammograms with (1) breast cancer with multiple lexicons such as mass, asymmetry, or architectural distortion without calcifications; (2) hidden calcifications that were difficult to mark; or (3) incomplete medical records. Results: A graph-convolutional-network-based model was developed. A total of 581 mammographic images from 292 cases of breast cancer were divided based on the calcification distribution pattern: diffuse (n = 67), regional (n = 115), group (n = 337), linear (n = 8), or segmental (n = 54). The classification performances were measured using metrics including precision, recall, F1 score, accuracy, and multi-class area under the receiver operating characteristic curve. The proposed model achieved a precision of 0.522 ± 0.028, sensitivity of 0.643 ± 0.017, specificity of 0.847 ± 0.009, F1 score of 0.559 ± 0.018, accuracy of 64.325 ± 1.694%, and area under the curve of 0.745 ± 0.030; thus, the method was found to be superior compared to all baseline models. The predicted linear and diffuse classifications were highly similar to the ground truth, and the predicted grouped and regional classifications were also superior compared to baseline models. The prediction results are interpretable using visualization methods to highlight the important calcification nodes in graphs. Conclusions: The proposed deep neural network framework is an AI solution that automatically detects and classifies calcification distribution patterns on mammographic images highly suspected of showing breast cancers. Further study of the AI model in an actual clinical setting and additional data collection will improve its performance.

Automated Machine Learning (AutoML)-Derived Preconception Predictive Risk Model to Guide Early Intervention for Gestational Diabetes Mellitus.

The increasing prevalence of gestational diabetes mellitus (GDM) is contributing to the rising global burden of type 2 diabetes (T2D) and intergenerational cycle of chronic metabolic disorders. Primary lifestyle interventions to manage GDM, including second trimester dietary and exercise guidance, have met with limited success due to late implementation, poor adherence and generic guidelines. In this study, we aimed to build a preconception-based GDM predictor to enable early intervention. We also assessed the associations of top predictors with GDM and adverse birth outcomes. Our evolutionary algorithm-based automated machine learning (AutoML) model was implemented with data from 222 Asian multi-ethnic women in a preconception cohort study, Singapore Preconception Study of Long-Term Maternal and Child Outcomes (S-PRESTO). A stacked ensemble model with a gradient boosting classifier and linear support vector machine classifier (stochastic gradient descent training) was derived using genetic programming, achieving an excellent AUC of 0.93 based on four features (glycated hemoglobin A1c (HbA1c), mean arterial blood pressure, fasting insulin, triglycerides/HDL ratio). The results of multivariate logistic regression model showed that each 1 mmol/mol increase in preconception HbA1c was positively associated with increased risks of GDM (p = 0.001, odds ratio (95% CI) 1.34 (1.13-1.60)) and preterm birth (p = 0.011, odds ratio 1.63 (1.12-2.38)). Optimal control of preconception HbA1c may aid in preventing GDM and reducing the incidence of preterm birth. Our trained predictor has been deployed as a web application that can be easily employed in GDM intervention programs, prior to conception.

Analysis of Dual Combination Therapies Used in Treatment of Hypertension in a Multinational Cohort.

Importance: More than 1 billion adults have hypertension globally, of whom 70% cannot achieve their hypertension control goal with monotherapy alone. Data are lacking on clinical use patterns of dual combination therapies prescribed to patients who escalate from monotherapy. Objective: To investigate the most common dual combinations prescribed for treatment escalation in different countries and how treatment use varies by age, sex, and history of cardiovascular disease. Design, Setting, and Participants: This cohort study used data from 11 electronic health record databases that cover 118 million patients across 8 countries and regions between January 2000 and December 2019. Included participants were adult patients (ages ≥18 years) who newly initiated antihypertensive dual combination therapy after escalating from monotherapy. There were 2 databases included for 3 countries: the Iqvia Longitudinal Patient Database (LPD) Australia and Electronic Practice-based Research Network 2019 linked data set from South Western Sydney Local Health District (ePBRN SWSLHD) from Australia, Ajou University School of Medicine (AUSOM) and Kyung Hee University Hospital (KHMC) databases from South Korea, and Khoo Teck Puat Hospital (KTPH) and National University Hospital (NUH) databases from Singapore. Data were analyzed from June 2020 through August 2021. Exposures: Treatment with dual combinations of the 4 most commonly used antihypertensive drug classes (angiotensin-converting enzyme inhibitor [ACEI] or angiotensin receptor blocker [ARB]; calcium channel blocker [CCB]; ß-blocker; and thiazide or thiazide-like diuretic). Main Outcomes and Measures: The proportion of patients receiving each dual combination regimen, overall and by country and demographic subgroup. Results: Among 970 335 patients with hypertension who newly initiated dual combination therapy included in the final analysis, there were 11 494 patients from Australia (including 9291 patients in Australia LPD and 2203 patients in ePBRN SWSLHD), 6980 patients from South Korea (including 6029 patients in Ajou University and 951 patients in KHMC), 2096 patients from Singapore (including 842 patients in KTPH and 1254 patients in NUH), 7008 patients from China, 8544 patients from Taiwan, 103 994 patients from France, 76 082 patients from Italy, and 754 137 patients from the US. The mean (SD) age ranged from 57.6 (14.8) years in China to 67.7 (15.9) years in the Singapore KTPH database, and the proportion of patients by sex ranged from 24 358 (36.9%) women in Italy to 408 964 (54.3%) women in the US. Among 12 dual combinations of antihypertensive drug classes commonly used, there were significant variations in use across country and patient subgroup. For example starting an ACEI or ARB monotherapy followed by a CCB (ie, ACEI or ARB + CCB) was the most commonly prescribed combination in Australia (698 patients in ePBRN SWSLHD [31.7%] and 3842 patients in Australia LPD [41.4%]) and Singapore (216 patients in KTPH [25.7%] and 439 patients in NUH [35.0%]), while in South Korea, CCB + ACEI or ARB (191 patients in KHMC [20.1%] and 1487 patients in Ajou University [24.7%]), CCB + ß-blocker (814 patients in Ajou University [13.5%] and 217 patients in KHMC [22.8%]), and ACEI or ARB + CCB (147 patients in KHMC [15.5%] and 1216 patients in Ajou University [20.2%]) were the 3 most commonly prescribed combinations. The distribution of 12 dual combination therapies were significantly different by age and sex in almost all databases. For example, use of ACEI or ARB + CCB varied from 873 of 3737 patients ages 18 to 64 years (23.4%) to 343 of 2292 patients ages 65 years or older (15.0%) in South Korea's Ajou University database (P for database distribution by age < .001), while use of ACEI or ARB + CCB varied from 2121 of 4718 (44.8%) men to 1721 of 4549 (37.7%) women in Australian LPD (P for drug combination distributions by sex < .001). Conclusions and Relevance: In this study, large variation in the transition between monotherapy and dual combination therapy for hypertension was observed across countries and by demographic group. These findings suggest that future research may be needed to investigate what dual combinations are associated with best outcomes for which patients.

Effect of Training Data Volume on Performance of Convolutional Neural Network Pneumothorax Classifiers.

Large datasets with high-quality labels required to train deep neural networks are challenging to obtain in the radiology domain. This work investigates the effect of training dataset size on the performance of deep learning classifiers, focusing on chest radiograph pneumothorax detection as a proxy visual task in the radiology domain. Two open-source datasets (ChestX-ray14 and CheXpert) comprising 291,454 images were merged and convolutional neural networks trained with stepwise increase in training dataset sizes. Model iterations at each dataset volume were evaluated on an external test set of 525 emergency department chest radiographs. Learning curve analysis was performed to fit the observed AUCs for all models generated. For all three network architectures tested, model AUCs and accuracy increased rapidly from 2 × 103 to 20 × 103 training samples, with more gradual increase until the maximum training dataset size of 291 × 103 images. AUCs for models trained with the maximum tested dataset size of 291 × 103 images were significantly higher than models trained with 20 × 103 images: ResNet-50: AUC20k = 0.86, AUC291k = 0.95, p < 0.001; DenseNet-121 AUC20k = 0.85, AUC291k = 0.93, p < 0.001; EfficientNet AUC20k = 0.92, AUC 291 k = 0.98, p < 0.001. Our study established learning curves describing the relationship between dataset training size and model performance of deep learning convolutional neural networks applied to a typical radiology binary classification task. These curves suggest a point of diminishing performance returns for increasing training data volumes, which algorithm developers should consider given the high costs of obtaining and labelling radiology data.