Best-performing Algorithm Research Articles

Smart predictions of petrophysical formation pore pressure via robust data-driven intelligent models

AbstractPredicting pore pressure in the formation is crucial for assessing reservoir geomechanical characteristics, designing drilling schemes/mud programs, and strategies to enhance oil recovery. Accurate predictions are vital for safe and cost-effective exploration and development. Recent research has seen the emergence of intelligent models utilizing machine learning (ML) and deep learning (DL) algorithms, offering promising outcomes. However, there remains a need to identify the most accurate and dependable model among these. This study aims to address this gap by comparing the performance of various ML and DL models, as reported in existing literature, to determine the optimal approach for pore pressure prediction. The sorted machine learning (ML) and deep learning (DL) regression algorithms used for the comparative analysis are decision tree (DT), extreme gradient boosting (XGBoost), random forest (RF), recurrent neural network (RNN), and convolutional neural network (CNN). A total dataset of 22,539 is gathered from five wells (15/9-F-1 A, 15/9-F-1 B, 15/9-F-11 A, 15/9-F-11 T2, and 15/9-F-14) drilled at North-sea Volve oil field, Norway. The first four wells are used to train and test the ML and DL algorithm, and the remaining well (15/9-F-14) is used to evaluate the best-performing algorithm’s universality in predicting pore pressure at the field of study. Seven different petrophysical parameters are used as input parameters to develop the predictive models. Statistical performance metrics are carried out to analyze the applied ML and DL performance. Based on performance indicators, the RF algorithm showed superior results compared to other predictive models with R2 and RMSE values of 0.97 and 2.70 MPa, respectively. Furthermore, the best-performing predictive model with low prediction error RMSE value is applied to the other well dataset from the field of study to access the universality of the RF algorithm to predict pore pressure in the field of study. The results of the universality analysis show a satisfactory prediction accuracy with R2 and RMSE values of 0.905 and 6.48 MPa, respectively.

One-Year-Old Precocious Chinese Mitten Crab Identification Algorithm Based on Task Alignment.

The cultivation of the Chinese mitten crab (Eriocheir sinensis) is an important component of China's aquaculture industry and also a field of concern worldwide. It focuses on the selection of high-quality, disease-free juvenile crabs. However, the early maturity rate of more than 18.2% and the mortality rate of more than 60% make it difficult to select suitable juveniles for adult culture. The juveniles exhibit subtle distinguishing features, and the methods for differentiating between sexes vary significantly; without training from professional breeders, it is challenging for laypersons to identify and select the appropriate juveniles. Therefore, we propose a task-aligned detection algorithm for identifying one-year-old precocious Chinese mitten crabs, named R-TNET. Initially, the required images were obtained by capturing key frames, and then they were annotated and preprocessed by professionals to build a training dataset. Subsequently, the ResNeXt network was selected as the backbone feature extraction network, with Convolutional Block Attention Modules (CBAMs) and a Deformable Convolution Network (DCN) embedded in its residual blocks to enhance its capability to extract complex features. Adaptive spatial feature fusion (ASFF) was then integrated into the feature fusion network to preserve the detailed features of small targets such as one-year-old precocious Chinese mitten crab juveniles. Finally, based on the detection head proposed by task-aligned one-stage object detection, the parameters of its anchor alignment metric were adjusted to detect, locate, and classify the crab juveniles. The experimental results showed that this method achieves a mean average precision (mAP) of 88.78% and an F1-score of 97.89%. This exceeded the best-performing mainstream object detection algorithm, YOLOv7, by 4.17% in mAP and 1.77% in the F1-score. Ultimately, in practical application scenarios, the algorithm effectively identified one-year-old precocious Chinese mitten crabs, providing technical support for the automated selection of high-quality crab juveniles in the cultivation process, thereby promoting the rapid development of aquaculture and agricultural intelligence in China.

Artificial Neural Network Prediction of Mortality in Cancer Patients Presenting for Radiation Therapy at a Multisite Institution.

For many decades, the management of cancer has utilized radiation therapy, which continues to evolve with technology to improve patient outcomes. However, despite the standardization of treatment plans and the establishment of best clinical practices based on prospective, randomized trials and adherence to National Comprehensive Cancer Network (NCCN) guidelines, the outcomes from radiation therapy are highly variable and dependent on a number of factors, including patient demographics, tumor characteristics/histology, and treatment parameters. In this study, we attempt to use available patient data and treatment parameters at the time of radiation therapy to predict future outcomes using artificial intelligence (AI). Six thousand five hundred ninety-five cases of patients who completed radiation treatment were selected retrospectively and used to train artificial neural networks (ANNs) and baseline models (i.e., logistic regression, random forest, support vector machines [SVMs], gradient boosting [XGBoost]) for binary classification of mortality at multiple time points ranging from six months to five years post-treatment. A hyperparameter grid search was used to identify the optimal network architecture for each time point, using sensitivity as the primary outcome metric. The median age was 75 years (range: 2-102 years). There were 63.8% females and 36.1% males. The results indicate that ANNs were able to successfully perform binary mortality prediction with an accuracy greater than random chance and greater sensitivity than baseline models used. The best-performing algorithm was the ANN, which achieved a sensitivity of 83.00% ± 4.89% for five-year mortality. The neural network was able to achieve higher sensitivity than Logistic Regression, SVM Random Forest, and XGBoost across all output target variables, demonstrating the utility of a neural network model for mortality prediction on the provided dataset.

Amplitude and frequency modulation of EEG predicts Intraventricular hemorrhage in preterm infants

BackgroundIntraventricular hemorrhage (IVH) is a common and significant complication in premature infants. While cranial ultrasound is the golden standard for IVH detection, it may not identify lesions until hours or days after occurring, which limits early intervention. Predicting IVH in premature infants would be highly advantageous. Recent studies have shown that EEG data’s amplitude and frequency modulation features could offer predictive insights for neurological diseases in adults. MethodsTo investigate the association between IVH and EEG monitoring, a retrospective case-control study was conducted in preterm infants. All infants underwent amplitude integrated EEG monitoring for at least 3 days after birth. The study included 20 cases who had an IVH diagnosed on cranial ultrasound and had a negative ultrasound 24 h earlier, and 20 matched controls without IVH. Amplitude and frequency modulation features were extracted from single-channel EEG data, and various machine learning algorithms were evaluated to create a predictive model. ResultsCases had an average gestational age and birth weight of 26.4 weeks and 965 g, respectively. The best-performing algorithm was adaptive boosting. EEG data from 24 h before IVH detection proved predictive with an area under the receiver operating characteristic curve of 93 %, an accuracy of 91 %, and a Kappa value of 0.85. The most informative features were the slow varying instantaneous frequency and amplitude in the Delta frequency band. ConclusionAmplitude and frequency modulation features obtained from single-channel EEG signals in extremely preterm infants show promise for predicting IVH occurrence within 24 h before detection on cranial ultrasound.

Comparative Analysis of Logistic Regression, Gradient Boosted Trees, SVM, and Random Forest Algorithms for Prediction of Acute Kidney Injury Requiring Dialysis After Cardiac Surgery.

This study aimed to identify the best-performing algorithm for predicting Acute Kidney Injury (AKI) necessitating dialysis following cardiac surgery. The dataset encompassed patient data from a tertiary cardiothoracic center in Malaysia between 2011 and 2015, sourced from electronic health records. Extensive preprocessing and feature selection ensured data quality and relevance. Four machine learning algorithms were applied: Logistic Regression, Gradient Boosted Trees, Support Vector Machine, and Random Forest. The dataset was split into training and validation sets and the hyperparameters were tuned. Accuracy, Area Under the ROC Curve (AUC), precision, F-measure, sensitivity, and specificity were some of the evaluation criteria. Ethical guidelines for data use and patient privacy were rigorously followed throughout the study. With the highest accuracy (88.66%), AUC (94.61%), and sensitivity (91.30%), Gradient Boosted Trees emerged as the top performance. Random Forest displayed strong AUC (94.78%) and accuracy (87.39%). In contrast, the Support Vector Machine showed higher sensitivity (98.57%) with lower specificity (59.55%), but lower accuracy (79.02%) and precision (70.81%). Sensitivity (87.70%) and specificity (87.05%) were maintained in balance via Logistic Regression. These findings imply that Gradient Boosted Trees and Random Forest might be an effective method for identifying patients who would develop AKI following heart surgery. However specific goals, sensitivity/specificity trade-offs, and consideration of the practical ramifications should all be considered when choosing an algorithm.

Decision support for traffic safety: case of rear-end collision modelling

ABSTRACT This paper presents the decision problem of selecting a collision-warning activation algorithm for Rear-End Collision Alert Signal (RECAS) systems installed in motorcycles and other powered two-wheelers. Using Agent-Based Modelling, Discrete Event Simulation and System Dynamics, we have developed a hybrid simulation model of three scenarios of rear-end collisions to measure the performance of several algorithms found in the literature. The best-performing algorithm, by Hirst & Graham, has reduced the probability of collisions on average by 31.21% in comparison with the baseline scenario of using a standard brake light. Using this algorithm also resulted in up to 11.20% reduction in the estimated severity of injuries when a rear-end collision was not averted. As the Hirst & Graham algorithm was the most effective in all the tested emergency braking scenarios and across all the tested speed limits, we recommend this algorithm to be used in RECAS systems in powered two-wheelers.

Deep learning in sex estimation from photographed human mandible using the Human Osteological Research Collection

Sex estimation is a necessary part of forensic and osteological analyses of skeletal human remains in the construction of a biological profile. Several skeletal traits are sexually dimorphic and used for skeletal sex estimation. The human mandible and morphological traits therein have been long used for sex estimation, but the validity of using the mandible in this purpose has become a concern. In this study, we examined the potential of artificial intelligence (AI) and especially deep learning (DL) to provide accurate sex estimations from the mandible. We used 193 modern South African mandibles from the Human Osteological Research Collection (HORC) in the Sefako Makgatho Health Sciences university with known sex to conduct our study. All mandibles were photographed from the same angle and the photographs were analyzed with an open-source DL software. The best-performing DL algorithm estimated the sex of males with 100% accuracy and females with 76.9% accuracy. However, further studies with a higher number of specimens could provide more reliable validity for using AI when building the biological profile from skeletal remains.

Development and validation of a claims-based algorithm to identify patients with Neuromyelitis Optica Spectrum disorder

IntroductionNo validated algorithm exists to identify patients with neuromyelitis optica spectrum disorder (NMOSD) in healthcare claims data. We developed and tested the performance of a healthcare claims–based algorithm to identify patients with NMOSD. MethodsUsing medical record data of 101 adults with NMOSD, multiple sclerosis (MS), or myelin oligodendrocyte glycoprotein antibody–associated disease (MOGAD), we tested the sensitivity and specificity of claims-based algorithms developed through interviews with neurologists. We tested the best-performing algorithm's face validity using 2016–2019 data from IBM MarketScan Commercial and Medicare Supplemental databases. Demographics and clinical characteristics were reported. ResultsAlgorithm inclusion criteria were age ≥ 18 years and (≥1 NMO diagnosis [or ≥ 1 transverse myelitis (TM) and ≥ 1 optic neuritis (ON) diagnosis] and ≥ 1 NMOSD drug) or (≥2 NMO diagnoses ≥90 days apart). Exclusion criteria were MS diagnosis or use of MS-specific drug after last NMO diagnosis or NMOSD drug; sarcoidosis diagnosis after last NMO diagnosis; or use of ≥1 immune checkpoint inhibitor. In medical record billing data of 50 patients with NMOSD, 30 with MS, and 21 with MOGAD, the algorithm had 82.0% sensitivity and 70.6% specificity. When applied to healthcare claims data, demographic and clinical features of the identified cohort were similar to known demographics of NMOSD. ConclusionsThis clinically derived algorithm performed well in medical records. When tested in healthcare claims, demographics and clinical characteristics were consistent with previous clinical findings. This algorithm will enable a more accurate estimation of NMOSD disease burden using insurance claims datasets.

Artificial intelligence streamlines diagnosis and assessment of prognosis in Brugada syndrome: a systematic review and meta-analysis

Aim: The objective of this systematic review and meta-analysis was to determine the diagnostic and prognostic utility of artificial intelligence/machine learning (AI/ML) algorithms in Brugada Syndrome (BrS). Methods: A systematic review and meta-analysis of the literature was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines. MEDLINE, EMBASE, SCOPUS, and WEB OF SCIENCE databases were searched for relevant articles. Abstract and title screening, full-text review, and data extraction were conducted independently by two of the authors. Conflicts were resolved via discussion among authors. A risk-of-bias assessment was performed using the QUADAS-2 tool for diagnostic studies and the PROBAST tool for prognostic studies. Forest plots and the summary area under the receiver operating characteristic (SAUROC) curve were done in R. Results: A total of 12 papers were included in our study. Among the best-performing diagnostic algorithms from each study, the sensitivity and specificity ranged from 0.80 to 0.89 and 0.74 to 0.97, respectively. In overall studies, sensitivity was 0.845 ± 0.014 and specificity was 0.892 ± 0.062 using a random effects model. A pooled analysis of the summary area under the receiver operating characteristic curve (SAUROC) was 0.77 for diagnostic studies. Prognostic studies showed good performance as well, with the AUC of the best-performing prognostic algorithms ranging from 0.71 to 0.90. Conclusions: Overall, AI/ML algorithms had high diagnostic and prognostic accuracy. These results highlight the potential of AI/ML algorithms for the diagnosis and prognosis of BrS and permit a choice of the best-performing ML algorithms.

Financial constraints prediction to lead socio-economic development: An application of neural networks to the Italian market

This study applies a neural network framework to optimize the classification of firms and to predict their difficulties in collecting external financial resources in the short term. In detail, we adopt replicated bootstrapped algorithms optimized on sensitivity and specificity as error measures and we propose a comparative analysis to identify the best-performing one. According to our results, the Conjugate gradient backpropagation with Fletcher-Reeves updates (i.e., CGF) is the best-performing algorithm, with sensitivity equal to 74.41 % and specificity equal to 70.11 %. Then, we use this algorithm and its weights to provide a classification of the Italian manufacturing industry in 2019, identifying the geographical areas in which firms under financial constraints are located, as well as the most critical industrial sectors. Based on this evidence, and considering the implementation of a cohesion policy, we highlight interventions by policy makers to support firms’ access to the capital market, fostering their investments and the consequent socio-economic development.

Improved mapping of highland bamboo forests using Sentinel-2 time series and machine learning in Google Earth Engine

Recent advances in the application of spectral bands from satellite observations and machine learning algorithms (MLA) in the Google Earth Engine (GEE) cloud-computing platform have been demonstrated to enhance the accuracy of mapping forest resources. This study presents a novel method for mapping the natural distribution of highland bamboo (Oldeania alpina) using spectral bands and three machine learning algorithms, namely random forest, gradient tree boosting, and classification and regression tree. First, spectral bands, vegetation indices and textural features including contrast, entropy and inverse difference moment were derived from the Sentinel-2 elevation data. Second, observations were categorized as the dry season (December–February), short rainy season (March-May), main (long) rainy season (June-August), wet season (September–November) and annual composite. Third, the machine learning algorithms were tested using 1882 ground control points collected from field and high spatial resolution images. Finally, the best-performing machine learning algorithm was used to classify and map bamboo forests. The five land cover categories identified were bamboo stands, natural forest, other vegetation, non-vegetated areas and water bodies. The result shows that the random forest classifier is a robust algorithm with an overall accuracy of 94% considering all the annual composite’s spectral, vegetation and textural variables. The highland bamboo coverage (91 km2) in the Andracha district provided valuable insights. Bamboo stands were mostly distributed in the southern and southeastern parts of the district. Here, we show that image compositing and multiple input parameters using machine learning techniques can overcome challenges facing land cover classification, which can hinder accurate mapping of Afromontane forests. We conclude that the incorporation of vegetation indices and textural features in land cover classification increases accuracy of mapping natural highland bamboo coverage and distribution. The application of this new method is a promising prospect not only in Ethiopia but also in Afromontane forests elsewhere in Africa.

Boosting wisdom of the crowd for medical image annotation using training performance and task features.

A crucial bottleneck in medical artificial intelligence (AI) is high-quality labeled medical datasets. In this paper, we test a large variety of wisdom of the crowd algorithms to label medical images that were initially classified by individuals recruited through an app-based platform. Individuals classified skin lesions from the International Skin Lesion Challenge 2018 into 7 different categories. There was a large dispersion in the geographical location, experience, training, and performance of the recruited individuals. We tested several wisdom of the crowd algorithms of varying complexity from a simple unweighted average to more complex Bayesian models that account for individual patterns of errors. Using a switchboard analysis, we observe that the best-performing algorithms rely on selecting top performers, weighting decisions by training accuracy, and take into account the task environment. These algorithms far exceed expert performance. We conclude by discussing the implications of these approaches for the development of medical AI.

Prediction of gas–solid erosion wear of bionic surfaces based on machine learning and unimodal intelligent optimization algorithm

Solid particle erosion wear is an inevitable phenomenon in industrial production, with erosion removal mass serving as a crucial metric for assessing the wear rate per unit area on the impacted surface. Developing predictive models to estimate the degree of mass removal is crucial for effectively controlling, evaluating, and preventing severe damage resulting from solid particle erosion wear. In this study, we constructed a comprehensive dataset comprising smooth and bionic surfaces, encompassing inner, outer, and planar surfaces. The dataset was used in a multifactorial erosion experiment, considering adjustable erosion angles, solid particle incident gas velocities, and solid particle diameter sizes. Through visualization and analysis of the obtained dataset, we identified optimal scenarios for bionic surface erosion resistance, offering insights for structural design against bionic erosion resistance. Furthermore, we compared machine learning algorithms to address the prediction problem, resulting in the selection of the best-performing regression algorithm, SVR (Support Vector Regression). Additionally, we compared the performance of other advanced intelligent optimization algorithms using unimodal benchmark functions, finding that HHO (Harris Hawks Optimization) emerged as the optimal choice for unimodal optimization. Building on HHO, we developed the IHHO (Improved Harris Hawks Optimization)-SVR model, using experimental data from erosion tests as the training dataset. This model can predict gas–solid two-phase flow erosion patterns, encompassing various wall types, solid particle sizes, solid incident gas velocities, and impact angles. Due to its robustness and rapid prediction capabilities, the model is expected to serve as a cost-effective tool for predicting erosion removal mass in gas–solid two-phase flow scenarios.

Emergency department risk model: timely identification of patients for outpatient care coordination.

Major depressive disorder (MDD) is linked to a 61% increased risk of emergency department (ED) visits and frequent ED usage. Collaborative care management (CoCM) models target MDD treatment in primary care, but how best to prioritize patients for CoCM to prevent frequent ED utilization remains unclear. This study aimed to develop and validate a risk identification model to proactively detect patients with MDD in CoCM at high risk of frequent (≥ 3) EDvisits. This retrospective cohort study utilized electronic health records from Mayo Clinic's primary care system to develop and validate a machine learning-based risk identification model. The model predicts the likelihood of frequent ED visits among patients with MDD within a 12-month period. Data were collected from Mayo Clinic's primary care system between May 1, 2006, and December 19, 2018. Risk identification models were developed and validated using machine learning classifiers to estimate frequent ED visit risks over 12 months. The Shapley Additive Explanations model identified variables driving frequent EDvisits. The patient population had a mean (SD) age of 39.78 (16.66) years, with 30.3% being male and 6.1%experiencing frequent ED visits. The best-performing algorithm (elastic-net logistic regression) achieved an area under the curve of 0.79 (95% CI, 0.74-0.84), a sensitivity of 0.71 (95% CI, 0.57-0.82), and a specificity of 0.76 (95% CI, 0.64-0.85) in the development data set. In the validation data set, the best-performing algorithm (random forest) achieved an area under the curve of 0.79, a sensitivity of 0.83, and a specificity of 0.61. Significant variables included male gender, prior frequent ED visits, high Patient Health Questionnaire-9 score, low education level, unemployment, and use of multiple medications. The risk identification model has potential for clinical application in triaging primary care patients with MDD in CoCM, aiming to reduce future ED utilization.

Predicting noise-induced hearing loss with machine learning: the influence of tinnitus as a predictive factor.

This study aimed to determine which machine learning model is most suitable for predicting noise-induced hearing loss and the effect of tinnitus on the models' accuracy. Two hundred workers employed in a metal industry were selected for this study and tested using pure tone audiometry. Their occupational exposure histories were collected, analysed and used to create a dataset. Eighty per cent of the data collected was used to train six machine learning models and the remaining 20 per cent was used to test the models. Eight workers (40.5 per cent) had bilaterally normal hearing and 119 (59.5 per cent) had hearing loss. Tinnitus was the second most important indicator after age for noise-induced hearing loss. The support vector machine was the best-performing algorithm, with 90 per cent accuracy, 91 per cent F1 score, 95 per cent precision and 88 per cent recall. The use of tinnitus as a risk factor in the support vector machine model may increase the success of occupational health and safety programmes.

The predictive value of supervised machine learning models for insomnia symptoms through smartphone usage behavior

IntroductionDigital phenotyping can be an innovative and unobtrusive way to improve the detection of insomnia. This study explores the correlations between smartphone usage features (SUF) and insomnia symptoms and their predictive value for detecting insomnia symptoms. MethodsIn an observational study of a German convenience sample, the Insomnia Severity Index (ISI) and smartphone usage data (e.g., time the screen was active, longest time the screen was inactive in the night) for the previous 7 days were obtained. SUF (e.g., min, mean) were calculated from the smartphone usage data. Correlation analyses between the ISI and SUF were conducted. For the specification of the machine learning models (ML), 80 % of the data was allocated to training, 20 % to testing, and five-fold cross-validation was used. Six algorithms (support vector machine, XGBoost, Random Forest, k-Nearest-Neighbor, Naive Bayes, and Logistic Regressions) were specified to predict ISI scores ≥15. Results752 participants (51.1 % female, mean ISI = 10.23, mean age = 41.92) were included in the analyses. Small correlations between some of the SUF and insomnia symptoms were found. In the ML models, sensitivity was low, ranging from 0.05 to 0.27 in the testing subsample. Random Forest and Naive Bayes were the best-performing algorithms. Yet, their AUCs (0.57, 0.58 respectively) in the testing subsample indicated a low discrimination capacity. ConclusionsGiven the small magnitude of the correlations and low discrimination capacity of the ML models, SUFs, as measured in this study, do not appear to be sufficient for detecting insomnia symptoms. Further research is necessary to explore whether examining intra-individual variations and subpopulations or employing alternative smartphone sensors yields more promising outcomes.

-A machine learning model to predict surgical site infection after surgery of lower extremity fractures.

This study aimed to develop machine learning algorithms for identifying predictive factors associated with the risk of postoperative surgical site infection in patients with lower extremity fractures. A machine learning analysis was conducted on a dataset comprising 1,579 patients who underwent surgical fixation for lower extremity fractures to create a predictive model for risk stratification of postoperative surgical site infection. We evaluated different clinical and demographic variables to train four machine learning models (neural networks, boosted generalised linear model, naïve bayes, and penalised discriminant analysis). Performance was measured by the area under the curve score, Youdon's index and Brier score. A multivariate adaptive regression splines (MARS) was used to optimise predictor selection. The final model consisted of five predictors. (1) Operating room time, (2) ankle region, (3) open injury, (4) body mass index, and (5) age. The best-performing machine learning algorithm demonstrated a promising predictive performance, with an area under the ROC curve, Youdon's index, and Brier score of 77.8%, 62.5%, and 5.1%-5.6%, respectively. The proposed predictive model not only assists surgeons in determining high-risk factors for surgical site infections but also empowers patients to closely monitor these factors and take proactive measures to prevent complications. Furthermore, by considering the identified predictors, this model can serve as a reference for implementing preventive measures and reducing postoperative complications, ultimately enhancing patient outcomes. However, further investigations involving larger datasets and external validations are required to confirm the reliability and applicability of our model.

Evaluating algorithms for identifying incident Guillain-Barré Syndrome in Medicare fee-for-service claims

ObjectiveClaims data can be leveraged to study rare diseases such as Guillain-Barré Syndrome (GBS), a neurological autoimmune condition. It is difficult to accurately measure and distinguish true cases of disease with claims without a validated algorithm. Our objective was to identify the best-performing algorithm for identifying incident GBS cases in Medicare fee-for-service claims data using chart reviews as the gold standard. Study design and settingThis was a multi-center, single institution cohort study from 2015 to 2019 that used Medicare-linked electronic health record (EHR) data. We identified 211 patients with a GBS diagnosis code in any position of an inpatient or outpatient claim in Medicare that also had a record of GBS in their electronic medical record. We reported the positive predictive value (PPV = number of true GBS cases/total number of GBS cases identified by the algorithm) for each algorithm tested. We also tested algorithms using several prevalence assumptions for false negative GBS cases and calculated a ranked sum for each algorithm's performance. ResultsWe found that 40 patients out of 211 had a true case of GBS. Algorithm 17, a GBS diagnosis in the primary position of an inpatient claim and a diagnostic procedure within 45 days of the inpatient admission date, had the highest PPV (PPV = 81.6%, 95% CI (69.3, 93.9). Across three prevalence assumptions, Algorithm 15, a GBS diagnosis in the primary position of an inpatient claim, was favored (PPV = 79.5%, 95% CI (67.6, 91.5). ConclusionsOur findings demonstrate that patients with incident GBS can be accurately identified in Medicare claims with a chart-validated algorithm. Using large-scale administrative data to study GBS offers significant advantages over case reports and patient repositories with self-reported data, and may be a potential strategy for the study of other rare diseases.

The Contribution of Explainable Machine Learning Algorithms Using ROI-based Brain Surface Morphology Parameters in Distinguishing Early-onset Schizophrenia From Bipolar Disorder

Background/aimTo differentiate early-onset schizophrenia (EOS) from early-onset bipolar disorder (EBD) using surface-based morphometry measurements and brain volumes using machine learning (ML) algorithms. MethodHigh-resolution T1-weighted images were obtained to measure cortical thickness (CT), gyrification, gyrification index (GI), sulcal depth (SD), fractal dimension (FD), and brain volumes. After the feature selection step, ML classifiers were applied for each feature set and the combination of them. The SHapley Additive exPlanations (SHAP) technique was implemented to interpret the contribution of each feature. FindingsOne-hundred-forty-four adolescents (16.2±1.4 years, female=40%) with EOS (n=81) and EBD (n=63) were included. The Adaptive Boosting (AdaBoost) algorithm had the highest accuracy (82.75%) in the whole dataset that includes all variables from Destrieux atlas. The best-performing algorithms were K-Nearest Neighbors (KNN) for FD subset, Support Vector Machine (SVM) for SD subset, and AdaBoost for GI subset. The KNN algorithm had the highest accuracy (accuracy=79.31%) in the whole dataset from the Desikan-Killiany-Tourville atlas. ConclusionThis study demonstrates the use of ML in the differential diagnosis of EOS and EBD using surface-based morphometry measurements. Future studies could focus on multicenter data for the validation of these results.

Real-World Gait Detection Using a Wrist-Worn Inertial Sensor: Validation Study.

Wrist-worn inertial sensors are used in digital health for evaluating mobility in real-world environments. Preceding the estimation of spatiotemporal gait parameters within long-term recordings, gait detection is an important step to identify regions of interest where gait occurs, which requires robust algorithms due to the complexity of arm movements. While algorithms exist for other sensor positions, a comparative validation of algorithms applied to the wrist position on real-world data sets across different disease populations is missing. Furthermore, gait detection performance differences between the wrist and lower back position have not yet been explored but could yield valuable information regarding sensor position choice in clinical studies. The aim of this study was to validate gait sequence (GS) detection algorithms developed for the wrist position against reference data acquired in a real-world context. In addition, this study aimed to compare the performance of algorithms applied to the wrist position to those applied to lower back-worn inertial sensors. Participants with Parkinson disease, multiple sclerosis, proximal femoral fracture (hip fracture recovery), chronic obstructive pulmonary disease, and congestive heart failure and healthy older adults (N=83) were monitored for 2.5 hours in the real-world using inertial sensors on the wrist, lower back, and feet including pressure insoles and infrared distance sensors as reference. In total, 10 algorithms for wrist-based gait detection were validated against a multisensor reference system and compared to gait detection performance using lower back-worn inertial sensors. The best-performing GS detection algorithm for the wrist showed a mean (per disease group) sensitivity ranging between 0.55 (SD 0.29) and 0.81 (SD 0.09) and a mean (per disease group) specificity ranging between 0.95 (SD 0.06) and 0.98 (SD 0.02). The mean relative absolute error of estimated walking time ranged between 8.9% (SD 7.1%) and 32.7% (SD 19.2%) per disease group for this algorithm as compared to the reference system. Gait detection performance from the best algorithm applied to the wrist inertial sensors was lower than for the best algorithms applied to the lower back, which yielded mean sensitivity between 0.71 (SD 0.12) and 0.91 (SD 0.04), mean specificity between 0.96 (SD 0.03) and 0.99 (SD 0.01), and a mean relative absolute error of estimated walking time between 6.3% (SD 5.4%) and 23.5% (SD 13%). Performance was lower in disease groups with major gait impairments (eg, patients recovering from hip fracture) and for patients using bilateral walking aids. Algorithms applied to the wrist position can detect GSs with high performance in real-world environments. Those periods of interest in real-world recordings can facilitate gait parameter extraction and allow the quantification of gait duration distribution in everyday life. Our findings allow taking informed decisions on alternative positions for gait recording in clinical studies and public health. ISRCTN Registry 12246987; https://www.isrctn.com/ISRCTN12246987. RR2-10.1136/bmjopen-2021-050785.