Prognosticating Fetal Growth Restriction and Small for Gestational Age by Medical History.

This study aimed to develop and externally validate a prognostic prediction model for screening fetal growth restriction (FGR)/small for gestational age (SGA) using medical history. From a nationwide health insurance database (n=1,697,452), we retrospectively selected visits of 12-to-55-year-old females to healthcare providers. This study used machine learning (including deep learning) and 54 medical-history predictors. The best model was a deep-insight visible neural network (DI-VNN). It had area under the curve of receiver operating characteristics (AUROC) 0.742 (95% CI 0.734 to 0.750) and a sensitivity of 49.09% (95% CI 47.60% to 50.58% at with 95% specificity). Our model used medical history for screening FGR/SGA with moderate accuracy by DI-VNN. In future work, we will compare this model with those from systematically-reviewed, previous studies and evaluate if this model's usage impacts patient outcomes.

Low- and high-level information analyses of transcriptome connecting endometrial-decidua-placental origin of preeclampsia subtypes: A preliminary study.

BACKGROUND: Existing proposed pathogenesis for preeclampsia (PE) was only applied for early onset subtype and did not consider pre-pregnancy and competing risks. We aimed to decipher PE subtypes by identifying related transcriptome that represents endometrial maturation and histologic chorioamnionitis. METHODS: We utilized eight arrays of mRNA expression for discovery (n=289), and other eight arrays for validation (n=352). Differentially expressed genes (DEGs) were overlapped between those of: (1) healthy samples from endometrium, decidua, and placenta, and placenta samples under histologic chorioamnionitis; and (2) placenta samples for each of the subtypes. They were all possible combinations based on four axes: (1) pregnancy-induced hypertension; (2) placental dysfunction-related diseases (e.g., fetal growth restriction [FGR]); (3) onset; and (4) severity. RESULTS: The DEGs of endometrium at late-secretory phase, but none of decidua, significantly overlapped with those of any subtypes with: (1) early onset (p-values ≤0.008); (2) severe hypertension and proteinuria (p-values ≤0.042); or (3) chronic hypertension and/or severe PE with FGR (p-values ≤0.042). Although sharing the same subtypes whose DEGs with which significantly overlap, the gene regulation was mostly counter-expressed in placenta under chorioamnionitis (n=13/18, 72.22%; odds ratio [OR] upper bounds ≤0.21) but co-expressed in late-secretory endometrium (n=3/9, 66.67%; OR lower bounds ≥1.17). Neither the placental DEGs at first-nor second-trimester under normotensive pregnancy significantly overlapped with those under late-onset, severe PE without FGR. CONCLUSIONS: We identified the transcriptome of endometrial maturation in placental dysfunction that distinguished early- and late-onset PE, and indicated chorioamnionitis as a PE competing risk. This study implied a feasibility to develop and validate the pathogenesis models that include pre-pregnancy and competing risks to decide if it is needed to collect prospective data for PE starting from pre-pregnancy including chorioamnionitis information.

Symptom-based scoring technique by machine learning to predict COVID-19: a validation study.

BACKGROUND: Coronavirus disease 2019 (COVID-19) surges, such as that which occurred when omicron variants emerged, may overwhelm healthcare systems. To function properly, such systems should balance detection and workloads by improving referrals using simple yet precise and sensitive diagnostic predictions. A symptom-based scoring system was developed using machine learning for the general population, but no validation has occurred in healthcare settings. We aimed to validate a COVID-19 scoring system using self-reported symptoms, including loss of smell and taste as major indicators. METHODS: A cross-sectional study was conducted to evaluate medical records of patients admitted to Dr. Sardjito Hospital, Yogyakarta, Indonesia, from March 2020 to December 2021. Outcomes were defined by a reverse-transcription polymerase chain reaction (RT-PCR). We compared the symptom-based scoring system, as the index test, with antigen tests, antibody tests, and clinical judgements by primary care physicians. To validate use of the index test to improve referral, we evaluated positive predictive value (PPV) and sensitivity. RESULTS: After clinical judgement with a PPV of 61% (n = 327/530, 95% confidence interval [CI]: 60% to 62%), confirmation with the index test resulted in the highest PPV of 85% (n = 30/35, 95% CI: 83% to 87%) but the lowest sensitivity (n = 30/180, 17%, 95% CI: 15% to 19%). If this confirmation was defined by either positive predictive scoring or antigen tests, the PPV was 92% (n = 55/60, 95% CI: 90% to 94%). Meanwhile, the sensitivity was 88% (n = 55/62, 95% CI: 87% to 89%), which was higher than that when using only antigen tests (n = 29/41, 71%, 95% CI: 69% to 73%). CONCLUSIONS: The symptom-based COVID-19 predictive score was validated in healthcare settings for its precision and sensitivity. However, an impact study is needed to confirm if this can balance detection and workload for the next COVID-19 surge.

Human-guided deep learning with ante-hoc explainability by convolutional network from non-image data for pregnancy prognostication.

BACKGROUND AND OBJECTIVE: Deep learning is applied in medicine mostly due to its state-of-the-art performance for diagnostic imaging. Supervisory authorities also require the model to be explainable, but most explain the model after development (post hoc) instead of incorporating explanation into the design (ante hoc). This study aimed to demonstrate a human-guided deep learning with ante-hoc explainability by convolutional network from non-image data to develop, validate, and deploy a prognostic prediction model for PROM and an estimator of time of delivery using a nationwide health insurance database. METHODS: To guide modeling, we constructed and verified association diagrams respectively from literatures and electronic health records. Non-image data were transformed into meaningful images utilizing predictor-to-predictor similarities, harnessing the power of convolutional neural network mostly used for diagnostic imaging. The network architecture was also inferred from the similarities. RESULTS: This resulted the best model for prelabor rupture of membranes (n=883, 376) with the area under curves 0.73 (95% CI 0.72 to 0.75) and 0.70 (95% CI 0.69 to 0.71) respectively by internal and external validations, and outperformed previous models found by systematic review. It was explainable by knowledge-based diagrams and model representation. CONCLUSIONS: This allows prognostication with actionable insights for preventive medicine.

Questionnaire-free machine-learning method to predict depressive symptoms among community-dwelling older adults.

The 15-item Geriatric Depression Scale (GDS-15) is widely used to screen for depressive symptoms among older populations. This study aimed to develop and validate a questionnaire-free, machine-learning model as an alternative triage test for the GDS-15 among community-dwelling older adults. The best models were the random forest (RF) and deep-insight visible neural network by internal validation, but both performances were undifferentiated by external validation. The AUROC of the RF model was 0.619 (95% CI 0.610 to 0.627) for the external validation set with a non-local ethnic group. Our triage test can allow healthcare professionals to preliminarily screen for depressive symptoms in older adults without using a questionnaire. If the model shows positive results, then the GDS-15 can be used for follow-up measures. This preliminary screening will save a lot of time and energy for healthcare providers and older adults, especially those persons who are illiterate.

Blood biomarkers representing maternal-fetal interface tissues used to predict early-and late-onset preeclampsia but not COVID-19 infection.

Background: A well-known blood biomarker (soluble fms-like tyrosinase-1 [sFLT-1]) for preeclampsia, i.e., a pregnancy disorder, was found to predict severe COVID-19, including in males. True biomarker may be masked by more-abrupt changes related to endothelial instead of placental dysfunction. This study aimed to identify blood biomarkers that represent maternal-fetal interface tissues for predicting preeclampsia but not COVID-19 infection. Methods: The surrogate transcriptome of tissues was determined by that in maternal blood, utilizing four datasets (n = 1354) which were collected before the COVID-19 pandemic. Applying machine learning, a preeclampsia prediction model was chosen between those using blood transcriptome (differentially expressed genes [DEGs]) and the blood-derived surrogate for tissues. We selected the best predictive model by the area under the receiver operating characteristic (AUROC) using a dataset for developing the model, and well-replicated in datasets both with and without an intervention. To identify eligible blood biomarkers that predicted any-onset preeclampsia from the datasets but that were not positive in the COVID-19 dataset (n = 47), we compared several methods of predictor discovery: (1) the best prediction model; (2) gene sets of standard pipelines; and (3) a validated gene set for predicting any-onset preeclampsia during the pandemic (n = 404). We chose the most predictive biomarkers from the best method with the significantly largest number of discoveries by a permutation test. The biological relevance was justified by exploring and reanalyzing low- and high-level, multiomics information. Results: A prediction model using the surrogates developed for predicting any-onset preeclampsia (AUROC of 0.85, 95 % confidence interval [CI] 0.77 to 0.93) was the only that was well-replicated in an independent dataset with no intervention. No model was well-replicated in datasets with a vitamin D intervention. None of the blood biomarkers with high weights in the best model overlapped with blood DEGs. Blood biomarkers were transcripts of integrin-α5 (ITGA5), interferon regulatory factor-6 (IRF6), and P2X purinoreceptor-7 (P2RX7) from the prediction model, which was the only method that significantly discovered eligible blood biomarkers (n = 3/100 combinations, 3.0 %; P =.036). Most of the predicted events (73.70 %) among any-onset preeclampsia were cluster A as defined by ITGA5 (Z-score ≥ 1.1), but were only a minority (6.34 %) among positives in the COVID-19 dataset. The remaining were predicted events (26.30 %) among any-onset preeclampsia or those among COVID-19 infection (93.66 %) if IRF6 Z-score was ≥-0.73 (clusters B and C), in which none was the predicted events among either late-onset preeclampsia (LOPE) or COVID-19 infection if P2RX7 Z-score was <0.13 (cluster C). Greater proportions of predicted events among LOPE were cluster A (82.85 % vs 70.53 %) compared to early-onset preeclampsia (EOPE). The biological relevance by multiomics information explained the biomarker mechanism, polymicrobial infection in any-onset preeclampsia by ITGA5, viral co-infection in EOPE by ITGA5-IRF6, a shared prediction with COVID-19 infection by ITGA5-IRF6-P2RX7, and non-replicability in datasets with a vitamin D intervention by ITGA5. Conclusions: In a model that predicts preeclampsia but not COVID-19 infection, the important predictors were genes in maternal blood that were not extremely expressed, including the proposed blood biomarkers. The predictive performance and biological relevance should be validated in future experiments.

Lightweight deep neural networks for cholelithiasis and cholecystitis detection by point-of-care ultrasound.

BACKGROUND AND OBJECTIVE: Emergency physicians (EPs) frequently deal with abdominal pain, including that is caused by either gallstones or acute cholecystitis. Easy access and low cost justify point-of-care ultrasound (POCUS) use as a first-line test to detect these diseases; yet, the detection performance of POCUS by EPs is unreliable, causing misdiagnoses with serious impacts. This study aimed to develop a machine learning system to detect and localize gallstones and to detect acute cholecystitis by ultrasound (US) still images taken by physicians or technicians for preliminary diagnoses. METHODS: Abdominal US images (> 89,000) were collected from 2386 patients in a hospital database. We constructed training sets for gallstones with or without cholecystitis (N = 10,971) and cholecystitis with or without gallstones (N = 7348) as positives. Validation sets were also constructed for gallstones (N = 2664) and cholecystitis (N = 1919). We applied a single-shot multibox detector (SSD) and a feature pyramid network (FPN) to classify and localize objects using image features extracted by ResNet-50 for gallstones, and MobileNet V2 to classify cholecystitis. The deep learning models were pretrained using the COCO-2017 and ILSVRC-2012 datasets. RESULTS: Using the validation sets, the SSD-FPN-ResNet-50 and MobileNet V2 achieved areas under the receiver operating characteristics curve of 0.92 and 0.94, respectively. The inference speeds were 21 (47.6 frames per second, fps) and 7 ms (142.9 fps). CONCLUSIONS: A machine learning system was developed to detect and localize gallstones, and to detect cholecystitis, with acceptable discrimination and speed. This is the first study to develop this system for either gallstone or cholecystitis detection with absence or presence of each one. After clinical trials, this system may be used to assist EPs, including those in remote areas, for detecting these diseases.

Deep learning for abdominal ultrasound: A computer-aided diagnostic system for the severity of fatty liver.

BACKGROUND: The prevalence of nonalcoholic fatty liver disease is increasing over time worldwide, with similar trends to those of diabetes and obesity. A liver biopsy, the gold standard of diagnosis, is not favored due to its invasiveness. Meanwhile, noninvasive evaluation methods of fatty liver are still either very expensive or demonstrate poor diagnostic performances, thus, limiting their applications. We developed neural network-based models to assess fatty liver and classify the severity using B-mode ultrasound (US) images. METHODS: We followed standards for reporting of diagnostic accuracy guidelines to report this study. In this retrospective study, we utilized B-mode US images from a consecutive series of patients to develop four-class, two-class, and three-class diagnostic prediction models. The images were eligible if confirmed by at least two gastroenterologists. We compared pretrained convolutional neural network models, consisting of visual geometry group (VGG)19, ResNet-50 v2, MobileNet v2, Xception, and Inception v2. For validation, we utilized 20% of the dataset resulting in >100 images for each severity category. RESULTS: There were 21,855 images from 2,070 patients classified as normal (N = 11,307), mild (N = 4,467), moderate (N = 3,155), or severe steatosis (N = 2,926). We used ResNet-50 v2 for the final model as the best ones. The areas under the receiver operating characteristic curves were 0.974 (mild steatosis vs others), 0.971 (moderate steatosis vs others), 0.981 (severe steatosis vs others), 0.985 (any severity vs normal), and 0.996 (moderate-to-severe steatosis/clinically abnormal vs normal-to-mild steatosis/clinically normal). CONCLUSION: Our deep learning models achieved comparable predictive performances to the most accurate, yet expensive, noninvasive diagnostic methods for fatty liver. Because of the discriminative ability, including for mild steatosis, significant impacts on clinical applications for fatty liver are expected. However, we need to overcome machine-dependent variation, motion artifacts, lacking of second confirmation from any other tools, and hospital-dependent regional bias.

Comparison of Multivariable Logistic Regression and Other Machine Learning Algorithms for Prognostic Prediction Studies in Pregnancy Care: Systematic Review and Meta-Analysis.

BACKGROUND: Predictions in pregnancy care are complex because of interactions among multiple factors. Hence, pregnancy outcomes are not easily predicted by a single predictor using only one algorithm or modeling method. OBJECTIVE: This study aims to review and compare the predictive performances between logistic regression (LR) and other machine learning algorithms for developing or validating a multivariable prognostic prediction model for pregnancy care to inform clinicians' decision making. METHODS: Research articles from MEDLINE, Scopus, Web of Science, and Google Scholar were reviewed following several guidelines for a prognostic prediction study, including a risk of bias (ROB) assessment. We report the results based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Studies were primarily framed as PICOTS (population, index, comparator, outcomes, timing, and setting): Population: men or women in procreative management, pregnant women, and fetuses or newborns; Index: multivariable prognostic prediction models using non-LR algorithms for risk classification to inform clinicians' decision making; Comparator: the models applying an LR; Outcomes: pregnancy-related outcomes of procreation or pregnancy outcomes for pregnant women and fetuses or newborns; Timing: pre-, inter-, and peripregnancy periods (predictors), at the pregnancy, delivery, and either puerperal or neonatal period (outcome), and either short- or long-term prognoses (time interval); and Setting: primary care or hospital. The results were synthesized by reporting study characteristics and ROBs and by random effects modeling of the difference of the logit area under the receiver operating characteristic curve of each non-LR model compared with the LR model for the same pregnancy outcomes. We also reported between-study heterogeneity by using τ2 and I2. RESULTS: Of the 2093 records, we included 142 studies for the systematic review and 62 studies for a meta-analysis. Most prediction models used LR (92/142, 64.8%) and artificial neural networks (20/142, 14.1%) among non-LR algorithms. Only 16.9% (24/142) of studies had a low ROB. A total of 2 non-LR algorithms from low ROB studies significantly outperformed LR. The first algorithm was a random forest for preterm delivery (logit AUROC 2.51, 95% CI 1.49-3.53; I2=86%; τ2=0.77) and pre-eclampsia (logit AUROC 1.2, 95% CI 0.72-1.67; I2=75%; τ2=0.09). The second algorithm was gradient boosting for cesarean section (logit AUROC 2.26, 95% CI 1.39-3.13; I2=75%; τ2=0.43) and gestational diabetes (logit AUROC 1.03, 95% CI 0.69-1.37; I2=83%; τ2=0.07). CONCLUSIONS: Prediction models with the best performances across studies were not necessarily those that used LR but also used random forest and gradient boosting that also performed well. We recommend a reanalysis of existing LR models for several pregnancy outcomes by comparing them with those algorithms that apply standard guidelines. TRIAL REGISTRATION: PROSPERO (International Prospective Register of Systematic Reviews) CRD42019136106; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=136106.

Artificial intelligence-assisted prediction of preeclampsia: Development and external validation of a nationwide health insurance dataset of the BPJS Kesehatan in Indonesia.

BACKGROUND: We developed and validated an artificial intelligence (AI)-assisted prediction of preeclampsia applied to a nationwide health insurance dataset in Indonesia. METHODS: The BPJS Kesehatan dataset have been preprocessed using a nested case-control design into preeclampsia/eclampsia (n = 3318) and normotensive pregnant women (n = 19,883) from all women with one pregnancy. The dataset provided 95 features consisting of demographic variables and medical histories started from 24 months to event and ended by delivery as the event. Six algorithms were compared by area under the receiver operating characteristics curve (AUROC) with a subgroup analysis by time to the event. We compared our model to similar prediction models from systematically reviewed studies. In addition, we conducted a text mining analysis based on natural language processing techniques to interpret our modeling results. FINDINGS: The best model consisted of 17 predictors extracted by a random forest algorithm. Nine∼12 months to the event was the period that had the best AUROC in external validation by either geographical (0.88, 95% confidence interval (CI) 0.88-0.89) or temporal split (0.86, 95% CI 0.85-0.86). We compared this model to prediction models in seven studies from 869 records in PUBMED, EMBASE, and SCOPUS. This model outperformed the previous models in terms of the precision, sensitivity, and specificity in all validation sets. INTERPRETATION: Our low-cost model improved preliminary prediction to decide pregnant women that will be predicted by the models with high specificity and advanced predictors. FUNDING: This work was supported by grant no. MOST108-2221-E-038-018 from the Ministry of Science and Technology of Taiwan.

Prediction of Preeclampsia and Intrauterine Growth Restriction: Development of Machine Learning Models on a Prospective Cohort.

BACKGROUND: Preeclampsia and intrauterine growth restriction are placental dysfunction-related disorders (PDDs) that require a referral decision be made within a certain time period. An appropriate prediction model should be developed for these diseases. However, previous models did not demonstrate robust performances and/or they were developed from datasets with highly imbalanced classes. OBJECTIVE: In this study, we developed a predictive model of PDDs by machine learning that uses features at 24-37 weeks' gestation, including maternal characteristics, uterine artery (UtA) Doppler measures, soluble fms-like tyrosine kinase receptor-1 (sFlt-1), and placental growth factor (PlGF). METHODS: A public dataset was taken from a prospective cohort study that included pregnant women with PDDs (66/95, 69%) and a control group (29/95, 31%). Preliminary selection of features was based on a statistical analysis using SAS 9.4 (SAS Institute). We used Weka (Waikato Environment for Knowledge Analysis) 3.8.3 (The University of Waikato, Hamilton, NZ) to automatically select the best model using its optimization algorithm. We also manually selected the best of 23 white-box models. Models, including those from recent studies, were also compared by interval estimation of evaluation metrics. We used the Matthew correlation coefficient (MCC) as the main metric. It is not overoptimistic to evaluate the performance of a prediction model developed from a dataset with a class imbalance. Repeated 10-fold cross-validation was applied. RESULTS: The classification via regression model was chosen as the best model. Our model had a robust MCC (.93, 95% CI .87-1.00, vs .64, 95% CI .57-.71) and specificity (100%, 95% CI 100-100, vs 90%, 95% CI 90-90) compared to each metric of the best models from recent studies. The sensitivity of this model was not inferior (95%, 95% CI 91-100, vs 100%, 95% CI 92-100). The area under the receiver operating characteristic curve was also competitive (0.970, 95% CI 0.966-0.974, vs 0.987, 95% CI 0.980-0.994). Features in the best model were maternal weight, BMI, pulsatility index of the UtA, sFlt-1, and PlGF. The most important feature was the sFlt-1/PlGF ratio. This model used an M5P algorithm consisting of a decision tree and four linear models with different thresholds. Our study was also better than the best ones among recent studies in terms of the class balance and the size of the case class (66/95, 69%, vs 27/239, 11.3%). CONCLUSIONS: Our model had a robust predictive performance. It was also developed to deal with the problem of a class imbalance. In the context of clinical management, this model may improve maternal mortality and neonatal morbidity and reduce health care costs.