Good Generalizability Research Articles

Exploring Inherent Consistency for Semi-Supervised Anatomical Structure Segmentation in Medical Imaging.

Due to the exorbitant expense of obtaining labeled data in the field of medical image analysis, semi-supervised learning has emerged as a favorable method for the segmentation of anatomical structures. Although semi-supervised learning techniques have shown great potential in this field, existing methods only utilize image-level spatial consistency to impose unsupervised regularization on data in label space. Considering that anatomical structures often possess inherent anatomical properties that have not been focused on in previous works, this study introduces the inherent consistency into semi-supervised anatomical structure segmentation. First, the prediction and the ground-truth are projected into an embedding space to obtain latent representations that encapsulate the inherent anatomical properties of the structures. Then, two inherent consistency constraints are designed to leverage these inherent properties by aligning these latent representations. The proposed method is plug-and-play and can be seamlessly integrated with existing methods, thereby collaborating to improve segmentation performance and enhance the anatomical plausibility of the results. To evaluate the effectiveness of the proposed method, experiments are conducted on three public datasets (ACDC, LA, and Pancreas). Extensive experimental results demonstrate that the proposed method exhibits good generalizability and outperforms several state-of-the-art methods.

Joint suppression of cardiac bSSFP cine banding and flow artifacts using twofold phase-cycling and a dual-encoder neural network

BackgroundCardiac bSSFP cine imaging suffers from banding and flow artifacts induced by off-resonance. The work aimed to develop a twofold phase cycling sequence with a neural network-based reconstruction (2P-SSFP+Network) for a joint suppression of banding and flow artifacts in cardiac cine imaging. MethodsA dual-encoder neural network was trained on 1620 pairs of phase-cycled left ventricular (LV) cine images collected from 18 healthy subjects. Twenty healthy subjects and 25 patients were prospectively scanned using the proposed 2P-SSFP sequence. bSSFP cine of a single RF phase increment (1P-SSFP), bSSFP cine of a single RF phase increment with a network-based artifact reduction (1P-SSFP+Network), the averaging of the two phase-cycled images (2P-SSFP+Average), and the proposed method were mutually compared, in terms of artifact suppression performance in the LV, generalizability over altered scan parameters and scanners, suppression of large-area banding artifacts in the left atrium (LA), and accuracy of downstream segmentation tasks. ResultsIn the healthy subjects, 2P-SSFP+Network showed robust suppressions of artifacts across a range of phase combinations. Compared with 1P-SSFP and 2P-SSFP+Average, 2P-SSFP+Network improved banding artifacts (3.85±0.67 and 4.50±0.45 vs 5.00±0.00, P<0.01 and P=0.02, respectively), flow artifacts (3.35±0.78 and 2.10±0.77 vs 4.90±0.20, both P<0.01), and overall image quality (3.25±0.51 and 2.30±0.60 vs 4.75±0.25, both P<0.01). 1P-SSFP+Network and 2P-SSFP+Network achieved a similar artifact suppression performance, yet the latter had fewer hallucinations (two-chamber, 4.25±0.51 vs 4.85±0.45, P=0.04; four-chamber, 3.45±1.21 vs 4.65±0.50, P=0.03; and LA, 3.35±1.00 vs 4.65±0.45, P<0.01). Furthermore, in the pulmonary veins and LA, 1P-SSFP+Network could not eliminate banding artifacts since they occupied a large area, whereas 2P-SSFP+Network reliably suppressed the artifacts. In the downstream automated myocardial segmentation task, 2P-SSFP+Network achieved more accurate segmentations than 1P-SSFP with different phase increments. Conclusions2P-SSFP+Network jointly suppresses banding and flow artifacts while manifesting a good generalizability against variations of anatomy and scan parameters. It provides a feasible solution for robust suppression of the two types of artifacts in bSSFP cine imaging.

MosquitoSong+: A noise-robust deep learning model for mosquito classification from wingbeat sounds.

In order to assess risk of mosquito-vector borne disease and to effectively target and monitor vector control efforts, accurate information about mosquito vector population densities is needed. The traditional and still most common approach to this involves the use of traps along with manual counting and classification of mosquito species, but the costly and labor-intensive nature of this approach limits its widespread use. Numerous previous studies have sought to address this problem by developing machine learning models to automatically identify species and sex of mosquitoes based on their wingbeat sounds. Yet little work has addressed the issue of robust classification in the presence of environmental background noise, which is essential to making the approach practical. In this paper, we propose a new deep learning model, MosquitoSong+, to identify the species and sex of mosquitoes from raw wingbeat sounds so that it is robust to the environmental noise and the relative volume of the mosquito's flight tone. The proposed model extends the existing 1D-CNN model by adjusting its architecture and introducing two data augmentation techniques during model training: noise augmentation and wingbeat volume variation. Experiments show that the new model has very good generalizability, with species classification accuracy above 80% on several wingbeat datasets with various background noise. It also has an accuracy of 93.3% for species and sex classification on wingbeat sounds overlaid with various background noises. These results suggest that the proposed approach may be a practical means to develop classification models that can perform well in the field.

A robust method for mapping soybean by phenological aligning of Sentinel-2 time series

Soybean is an important crop for food and animal feed. Production and area both continue to increase and expand into new areas and countries. Spatially explicit information on soybean cultivation is essential to crop monitoring, production estimation, and national accounting systems. However, its cultivation in diverse climate conditions, landscapes, and agricultural systems poses challenges to accurately map soybean across different regions and years. We propose an innovative soybean mapping method combining phenological alignment with machine learning (named here RF-DTW), which can be applied to diverse geographies and years by aligning phenological shifts and using distinctive features from Sentinel-2 time-series. The method first uses the dynamic time warping (DTW) algorithm to align the growing season between pixels across different sites. Then, based on the harmonized time-series, a set of distinctive features was identified and used to build random forest (RF) models to classify soybean across ten globally distributed sites and multiple years. Results show that the green chlorophyll vegetation index (GCVI), greenness and water content composite index (GWCCI), normalized difference senescent vegetation index (NDSVI), red edge position (REP), and short-wave infrared bands are important inputs for distinguishing soybean from other crops. Spectral-phenological features, particularly the curve slope metrics of GCVI and GWCCI during the peak to late growing season, rank as the most important features for mapping soybean. RF-DTW demonstrates good generalizability across ten study sites, achieving an overall accuracy (OA) of 0.92 and an F1-score of 0.84. F1-scores for eight out of ten sites ranged between 0.82 and 0.98, outperforming a benchmark method, although they were lower (F1-score < 0.60) for the two sites in Sub-Saharan Africa. Additionally, RF-DTW performs robustly when transferred to untrained regions and years, with most cases showing an F1-score higher than 0.70. Our proposed method, as a combination of phenological alignment and machine learning, can be used to map soybean accurately and efficiently across different regions and years, to provide crucial information for understanding the rapid dynamics of soybean cultivation and its global-scale impacts.

Explainable Machine-Learning Models to Predict Weekly Risk of Hyperglycemia, Hypoglycemia, and Glycemic Variability in Patients With Type 1 Diabetes Based on Continuous Glucose Monitoring.

The aim of this study was to develop and validate explainable prediction models based on continuous glucose monitoring (CGM) and baseline data to identify a week-to-week risk of CGM key metrics (hyperglycemia, hypoglycemia, glycemic variability). By having a weekly prediction of CGM key metrics, it is possible for the patient or health care personnel to take immediate preemptive action. We analyzed, trained, and internally tested three prediction models (Logistic regression, XGBoost, and TabNet) using CGM data from 187 type 1 diabetes patients with long-term CGM monitoring. A binary classification approach combined with feature engineering deployed on the CGM signals was used to predict hyperglycemia, hypoglycemia, and glycemic variability based on consensus targets (time above range ≥5%, time below range ≥4%, coefficient of variation ≥36%). The models were validated in two independent cohorts with a total of 223 additional patients of varying ages. A total of 46 593 weeks of CGM data were included in the analysis. For the best model (XGBoost), the area under the receiver operating characteristic curve (ROC-AUC) was 0.9 [95% confidence interval (CI) = 0.89-0.91], 0.89 [95% CI = 0.88-0.9], and 0.8 [95% CI = 0.79-0.81] for predicting hyperglycemia, hypoglycemia, and glycemic variability in the interval validation, respectively. The validation test showed good generalizability of the models with ROC-AUC of 0.88 to 0.95, 0.84 to 0.89, and 0.80 to 0.82 for predicting the glycemic outcomes. Prediction models based on real-world CGM data can be used to predict the risk of unstable glycemic control in the forthcoming week. The models showed good performance in both internal and external validation cohorts.

Remaining Useful Life of the Rolling Bearings Prediction Method Based on Transfer Learning Integrated with CNN-GRU-MHA

To accurately predict the remaining useful life (RUL) of rolling bearings under limited data and fluctuating load conditions, we propose a new method for constructing health indicators (HI) and a transfer learning prediction framework, which integrates Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Multi-head attention (MHA). Firstly, we combined Convolutional Neural Networks (CNN) with Gated Recurrent Units (GRU) to fully extract temporal and spatial features from vibration signals. Then, the Multi-head attention mechanism (MHA) was added for weighted processing to improve the expression ability of the model. Finally, a new method for constructing Health indicators (HIs) was proposed in which the noise reduction and normalized vibration signals were taken as a HI, the L1 regularization method was added to avoid overfitting, and the model-based transfer learning method was used to realize the RUL prediction of bearings under small samples and variable load conditions. Experiments were conducted using the PHM2012 dataset from the FEMTO-ST research institute and XJTU-SY dataset. Three sets of 12 migration experiments were conducted under three different operating conditions on the PHM2012 dataset. The results show that the average RMSE of the proposed method was 0.0443, indicating high prediction accuracy under variable loads and small sample conditions. Three different operating conditions and two sets of four migration experiments were conducted on the XJTU-SY dataset, and the results show that the average RMSE of the proposed method was 0.0693, verifying the good generalization of the model under variable load conditions. In summary, the proposed HI construction method and prediction framework can effectively reduce the differences between features, with high stability and good generalizability.

Novel multifactor predictive model for postoperative survival in gallbladder cancer: a multi-center study

BackgroundGallbladder cancer (GBC) is a highly aggressive malignancy, with limited survival profiles after curative surgeries. This study aimed to develop a practical model for predicting the postoperative overall survival (OS) in GBC patients.MethodsPatients from three hospitals were included. Two centers (N = 102 and 100) were adopted for model development and internal validation, and the third center (N = 85) was used for external testing. Univariate and stepwise multivariate Cox regression were used for feature selection. A nomogram for 1-, 3-, and 5-year postoperative survival rates was constructed accordingly. Performance assessment included Harrell's concordance index (C-index), receiver operating characteristic (ROC) curves and calibration curves. Kaplan-Meier curves were utilized to evaluate the risk stratification results of the nomogram. Decision curves were used to reflect the net benefit.ResultsEight factors, TNM stage, age-adjusted Charlson Comorbidity Index (aCCI), body mass index (BMI), R0 resection, blood platelet count, and serum levels of albumin, CA125, CA199 were incorporated in the nomogram. The time-dependent C-index consistently exceeded 0.70 from 6 months to 5 years, and time-dependent ROC revealed an area under the curve (AUC) of over 75% for 1-, 3-, and 5-year survival. The calibration curves, Kaplan-Meier curves and decision curves also indicated good prognostic performance and clinical benefit, surpassing traditional indicators TNM staging and CA199 levels. The reliability of results was further proved in the independent external testing set.ConclusionsThe novel nomogram exhibited good prognostic efficacy and robust generalizability in GBC patients, which might be a promising tool for aiding clinical decision-making.

Development of machine learning-based quantitative structure–activity relationship models for predicting plasma half-lives of drugs in six common food animal species

Abstract Plasma half-life is a crucial pharmacokinetic parameter for estimating extralabel withdrawal intervals of drugs to ensure the safety of food products derived from animals. This study focuses on developing a quantitative structure–activity relationship (QSAR) model incorporating multiple machine learning and artificial intelligence algorithms, and aims to predict the plasma half-lives of drugs in 6 food animals, including cattle, chickens, goats, sheep, swine, and turkeys. By integrating 4 machine learning algorithms with 5 molecular descriptor types, 20 QSAR models were developed using data from the Food Animal Residue Avoidance Databank (FARAD) Comparative Pharmacokinetic Database. The deep neural network (DNN) algorithm demonstrated the best prediction ability of plasma half-lives. The DNN model with all descriptors achieved superior performance with a high coefficient of determination (R2) of 0.82 ± 0.19 in 5-fold cross-validation on the training sets and an R2 of 0.67 on the independent test set, indicating accurate predictions and good generalizability. The final model was converted to a user-friendly web dashboard to facilitate its wide application by the scientific community. This machine learning-based QSAR model serves as a valuable tool for predicting drug plasma half-lives and extralabel withdrawal intervals in 6 common food animals based on physicochemical properties. It also provides a foundation to develop more advanced models to predict the tissue half-life of drugs in food animals.

Operative versus conservative management for inguinal hernia: a methodology scoping review of randomized controlled trials.

There is a lack of consensus on the management of inguinal hernia with limited symptoms. To address this issue a systematic review of existing randomized clinical trials (RCTs) was performed to critically appraise all existing data on asymptomatic hernia management, focusing on generalizability. A scoping review to identify all RCTs comparing surgical and conservative management of patients with inguinal hernias was undertaken. Medline, Embase, Cochrane and ClinicalTrials.gov databases were searched. Data collected included study characteristics and definitions of population, intervention/comparator, and outcomes; and limitations of each study were also extracted. The quality and generalizability of included RCTs were evaluated using Cochrane's ROB-2 and the PRECIS-2 tool, respectively. Searches returned 661 papers; 14 full-text papers were assessed and three RCTs were identified. All RCTs included only male patients with a mean age above 55 years. All RCTs included asymptomatic patients and two included those with minimal symptoms. Different definitions for 'minimally symptomatic' were used in RCTs and none provided details of what was meant by conservative treatment. Follow-up periods varied between studies (1, 2, 3 years). All RCTs had an overall high risk of bias. According to PRECIS-2, two RCTs were classified as pragmatic, and one was equally pragmatic and explanatory. This systematic review highlights a high risk of bias but a good generalizability of the findings from the RCTs conducted on minimally symptomatic inguinal hernia patients. To improve the guidelines for the management of this group of patients, more generalizable data are needed.

Predicting cognitive load with EEG using Riemannian geometry-based features

Objective. We show that electroencephalography (EEG)-based cognitive load (CL) prediction using Riemannian geometry features outperforms existing models. The performance is estimated using Riemannian Procrustes Analysis (RPA) with a test set of subjects unseen during training.Approach. Performance is evaluated by using the Minimum Distance to Riemannian Mean model trained on CL classification. The baseline performance is established using spatial covariance matrices of the signal as features. Various novel features are explored and analyzed in depth, including spatial covariance and correlation matrices computed on the EEG signal and its first-order derivative. Furthermore, each RPA step effect on the performance is investigated, and the generalization performance of RPA is compared against a few different generalization methods.Main results. Performances are greatly improved by using the spatial covariance matrix of the first-order derivative of the signal as features. Furthermore, this work highlights both the importance and efficiency of RPA for CL prediction: it achieves good generalizability with little amounts of calibration data and largely outperforms all the comparison methods.Significance. CL prediction using RPA for generalizability across subjects is an approach worth exploring further, especially for real-world applications where calibration time is limited. Furthermore, the feature exploration uncovers new, promising features that can be used and further experimented within any Riemannian geometry setting.

GPTFF: A high-accuracy out-of-the-box universal AI force field for arbitrary inorganic materials

This study introduces a novel artificial intelligence (AI) force field, namely a graph-based pre-trained transformer force field (GPTFF), which can simulate arbitrary inorganic systems with good precision and generalizability. Harnessing a large trove of the data and the attention mechanism of transformer algorithms, the model can accurately predict energy, atomic force, and stress with mean absolute error (MAE) values of 32 meV/atom, 71 meV/Å, and 0.365 GPa, respectively. The dataset used to train the model includes 37.8 million single-point energies, 11.7 billion force pairs, and 340.2 million stresses. We also demonstrated that the GPTFF can be universally used to simulate various physical systems, such as crystal structure optimization, phase transition simulations, and mass transport. The model is publicly released with this paper, enabling anyone to use it immediately without needing to train it.

Semi-supervised lung adenocarcinoma histopathology image classification based on multi-teacher knowledge distillation.

In this study, we propose a semi-supervised learning scheme using a patch-based deep learning framework to tackle the challenge of high-precision classification of seven lung tumor growth patterns, despite having a small amount of labeled data in whole slide images (WSIs). This scheme aims to enhance generalization ability with limited data and reduce dependence on large amounts of labeled data. It effectively addresses the common challenge of high demand for labeled data in medical image analysis.&#xD;&#xD;Approach. To address these challenges, the study employs a semi-supervised learning approach enhanced by a dynamic confidence threshold mechanism. This mechanism adjusts based on the quantity and quality of pseudo labels generated. This dynamic thresholding mechanism helps avoid the imbalance of pseudo-label categories and the low number of pseudo-labels that may result from a higher fixed threshold. Furthermore, the research introduces a multi-teacher knowledge distillation technique. This technique adaptively weights predictions from multiple teacher models to transfer reliable knowledge and safeguard student models from low-quality teacher predictions.&#xD;&#xD;Main results. The framework underwent rigorous training and evaluation using a dataset of 150 WSIs, each representing one of the seven growth patterns. The experimental results demonstrate that the framework is highly accurate in classifying lung tumor growth patterns in histopathology images. Notably, the performance of the framework is comparable to that of fully supervised models and human pathologists. In addition, the framework's evaluation metrics on a publicly available dataset are higher than those of previous studies, indicating good generalizability.&#xD;&#xD;Significance. This research demonstrates that a semi-supervised learning approach can achieve results comparable to fully supervised models and expert pathologists, thus opening new possibilities for efficient and cost-effective medical images analysis. The implementation of dynamic confidence thresholding and multi-teacher knowledge distillation techniques represents a significant advancement in applying deep learning to complex medical image analysis tasks. This advancement could lead to faster and more accurate diagnoses, ultimately improving patient outcomes and fostering the overall progress of healthcare technology.&#xD.

Interpretable semi-supervised clustering enables universal detection and intensity assessment of diverse aviation hazardous winds

The identification of aviation hazardous winds is crucial and challenging in air traffic management for assuring flight safety, particularly during the take-off and landing phases. Existing criteria are typically tailored for special wind types, and whether there exists a universal feature that can effectively detect diverse types of hazardous winds from radar/lidar observations remains as an open question. Here we propose an interpretable semi-supervised clustering paradigm to solve this problem, where the prior knowledge and probabilistic models of winds are integrated to overcome the bottleneck of scarce labels (pilot reports). Based on this paradigm, a set of high-dimensional hazard features is constructed to effectively identify the occurrence of diverse hazardous winds and assess the intensity metrics. Verification of the paradigm across various scenarios has highlighted its high adaptability to diverse input data and good generalizability to diverse geographical and climate zones.

Causal deconfounding deep reinforcement learning for mobile robot motion planning

Deep reinforcement learning (DRL) has emerged as an efficient approach for motion planning in mobile robot systems. It leverages the offline training process to enhance real-time computation efficiency. In DRL-based methods, the DRL models are trained to compute an action based on the current state of the robot and the surrounding obstacles. However, the trained models may capture spurious correlations through potential confounders, resulting in non-robust state representations, which limits the models’ robustness and generalizability. In this paper, we propose a Causal Deconfounding DRL method for Motion Planning, CD-DRL-MP, to address spurious correlations and learn robust and generalizable policies. Specifically, we formalize the temporal causal relationships between states and actions using a structural causal model. We then extract the minimal sufficient state representation set by blocking the backdoor paths in the causal model. Finally, using the representation set, CD-DRL-MP learns the causal effect between states and actions while mitigating the detrimental influence of potential confounders and computes motion commands for mobile robots. Comprehensive experiments show that the proposed method significantly outperforms non-causal DRL methods and existing causal methods, while guaranteeing good robustness and generalizability.

Machine Learning Prediction of Autism Spectrum Disorder From a Minimal Set of Medical and Background Information

Early identification of the likelihood of autism spectrum disorder (ASD) using minimal information is crucial for early diagnosis and intervention, which can affect developmental outcomes. To develop and validate a machine learning (ML) model for predicting ASD using a minimal set of features from background and medical information and to evaluate the predictors and the utility of the ML model. For this diagnostic study, a retrospective analysis of the Simons Foundation Powering Autism Research for Knowledge (SPARK) database, version 8 (released June 6, 2022), was conducted, including data from 30 660 participants after adjustments for missing values and class imbalances (15 330 with ASD and 15 330 without ASD). The SPARK database contains participants recruited from 31 university-affiliated research clinicals and online in 26 states in the US. All individuals with a professional ASD diagnosis and their families were eligible to participate. The model performance was validated on independent datasets from SPARK, version 10 (released July 21, 2023), and the Simons Simplex Collection (SSC), consisting of 14 790 participants, followed by phenotypic associations. Twenty-eight basic medical screening and background history items present before 24 months of age. Generalizable ML prediction models were developed for detecting ASD using 4 algorithms (logistic regression, decision tree, random forest, and eXtreme Gradient Boosting [XGBoost]). Performance metrics included accuracy, area under the receiver operating characteristics curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and F1 score, offering a comprehensive assessment of the predictive accuracy of the model. Explainable AI methods were applied to determine the effect of individual features in predicting ASD as secondary outcomes, enhancing the interpretability of the best-performing model. The secondary outcome analyses were further complemented by examining differences in various phenotypic measures using nonparametric statistical methods, providing insights into the ability of the model to differentiate between different presentations of ASD. The study included 19 477 (63.5%) male and 11 183 (36.5%) female participants (mean [SD] age, 106 [62] months). The mean (SD) age was 113 (68) months for the ASD group and 100 (55) months for the non-ASD group. The XGBoost (termed AutMedAI) model demonstrated strong performance with an AUROC score of 0.895, sensitivity of 0.805, specificity of 0.829, and PPV of 0.897. Developmental milestones and eating behavior were the most important predictors. Validation on independent cohorts showed an AUROC of 0.790, indicating good generalizability. In this diagnostic study of ML prediction of ASD, robust model performance was observed to identify autistic individuals with more symptoms and lower cognitive levels. The robustness and ML model generalizability results are promising for further validation and use in clinical and population settings.

Automated early detection of acute retinal necrosis from ultra-widefield color fundus photography using deep learning

BackgroundAcute retinal necrosis (ARN) is a relatively rare but highly damaging and potentially sight-threatening type of uveitis caused by infection with the human herpesvirus. Without timely diagnosis and appropriate treatment, ARN can lead to severe vision loss. We aimed to develop a deep learning framework to distinguish ARN from other types of intermediate, posterior, and panuveitis using ultra-widefield color fundus photography (UWFCFP).MethodsWe conducted a two-center retrospective discovery and validation study to develop and validate a deep learning model called DeepDrARN for automatic uveitis detection and differentiation of ARN from other uveitis types using 11,508 UWFCFPs from 1,112 participants. Model performance was evaluated with the area under the receiver operating characteristic curve (AUROC), the area under the precision and recall curves (AUPR), sensitivity and specificity, and compared with seven ophthalmologists.ResultsDeepDrARN for uveitis screening achieved an AUROC of 0.996 (95% CI: 0.994–0.999) in the internal validation cohort and demonstrated good generalizability with an AUROC of 0.973 (95% CI: 0.956–0.990) in the external validation cohort. DeepDrARN also demonstrated excellent predictive ability in distinguishing ARN from other types of uveitis with AUROCs of 0.960 (95% CI: 0.943–0.977) and 0.971 (95% CI: 0.956–0.986) in the internal and external validation cohorts. DeepDrARN was also tested in the differentiation of ARN, non-ARN uveitis (NAU) and normal subjects, with sensitivities of 88.9% and 78.7% and specificities of 93.8% and 89.1% in the internal and external validation cohorts, respectively. The performance of DeepDrARN is comparable to that of ophthalmologists and even exceeds the average accuracy of seven ophthalmologists, showing an improvement of 6.57% in uveitis screening and 11.14% in ARN identification.ConclusionsOur study demonstrates the feasibility of deep learning algorithms in enabling early detection, reducing treatment delays, and improving outcomes for ARN patients.

Key process identification method for aircraft skin milling based on quality characteristics and process correlation analysis

As a key aircraft component, the machining quality of aircraft skin is difficult to control. This paper proposes a key process identification method for aircraft skin milling, which comprehensively considers the influence of key quality features and the correlation between features and processes. The Fuzzy Analytic Hierarchy Process and Entropy Weight Method are used to calculate the influence weights of quality features, and the key quality features are comprehensively identified by the improved Fuzzy Borda Method. The improved algorithm reduces the non-linearity of calculation and the weakening of evaluation, solving the problem of low discrimination in the recognition results of a single evaluation method; At the same time, combining factors such as process difficulty, the relationship between quality characteristics and processes was analyzed, and key processes were identified by calculating relational grade. The experimental results show that the key processes of skin milling are the previous procedures of key quality features and the process located in the middle of the skin. The recognition results are more accurate and scientific compared to the recognition methods that do not consider the relationship between features and processes. This has certain guiding significance for practical aircraft skin machining tools and has good generalizability.

Causality-inspired Domain Expansion network for single domain generalization

Most existing single domain generalization (SDG) methods focus on extending the distribution of the training data to achieve good generalizability in the target domain by synthesizing out-of-domain data. However, in practice, the generated data are insufficient to cover all possible shifts. When the target distributions are not covered, the spurious associations between irrelevant features and labels captured by a model will deteriorate the generalization performance. To alleviate this problem, in this paper, we first analyze that augmenting the source domain with samples of new styles (e.g., different backgrounds) can mitigate spurious associations between irrelevant features and labels from a causal perspective, which can reduce the interference of non-causal features on the classification labels then propose a novel Causality-inspired Domain Expansion Network (CDEN) for SDG. Specifically, CDEN contains a domain augmentation subnetwork and a representation learning subnetwork. The former uses a special style generator module AdaIN∗ to generate out-of-domain data with the same semantic information but different styles from the source domain data to help the model capture true associations between discriminative features and labels. The latter introduces a novel class-level contrastive learning, instance-level contrastive learning, and invariant risk minimization to help the model further eliminate spurious associations. CDEN progressively generates diverse style data and learns domain invariant representations containing few spurious associations by optimizing the two subnetworks in an adversarial learning manner. We conduct extensive experiments on three benchmark datasets to evaluate the superiority of CDEN, comparing it to fifteen state-of-the-art methods.

Proteomic signatures improve risk prediction for common and rare diseases

For many diseases there are delays in diagnosis due to a lack of objective biomarkers for disease onset. Here, in 41,931 individuals from the United Kingdom Biobank Pharma Proteomics Project, we integrated measurements of ~3,000 plasma proteins with clinical information to derive sparse prediction models for the 10-year incidence of 218 common and rare diseases (81–6,038 cases). We then compared prediction models developed using proteomic data with models developed using either basic clinical information alone or clinical information combined with data from 37 clinical assays. The predictive performance of sparse models including as few as 5 to 20 proteins was superior to the performance of models developed using basic clinical information for 67 pathologically diverse diseases (median delta C-index = 0.07; range = 0.02–0.31). Sparse protein models further outperformed models developed using basic information combined with clinical assay data for 52 diseases, including multiple myeloma, non-Hodgkin lymphoma, motor neuron disease, pulmonary fibrosis and dilated cardiomyopathy. For multiple myeloma, single-cell RNA sequencing from bone marrow in newly diagnosed patients showed that four of the five predictor proteins were expressed specifically in plasma cells, consistent with the strong predictive power of these proteins. External replication of sparse protein models in the EPIC-Norfolk study showed good generalizability for prediction of the six diseases tested. These findings show that sparse plasma protein signatures, including both disease-specific proteins and protein predictors shared across several diseases, offer clinically useful prediction of common and rare diseases.

Cartesian atomic cluster expansion for machine learning interatomic potentials

Machine learning interatomic potentials are revolutionizing large-scale, accurate atomistic modeling in material science and chemistry. Many potentials use atomic cluster expansion or equivariant message-passing frameworks. Such frameworks typically use spherical harmonics as angular basis functions, followed by Clebsch-Gordan contraction to maintain rotational symmetry. We propose a mathematically equivalent and simple alternative that performs all operations in the Cartesian coordinates. This approach provides a complete set of polynormially independent features of atomic environments while maintaining interaction body orders. Additionally, we integrate low-dimensional embeddings of various chemical elements, trainable radial channel coupling, and inter-atomic message passing. The resulting potential, named Cartesian Atomic Cluster Expansion (CACE), exhibits good accuracy, stability, and generalizability. We validate its performance in diverse systems, including bulk water, small molecules, and 25-element high-entropy alloys.