A supervised contrastive learning method based on online complement strategy for long-tailed fine-grained fault diagnosis

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 64, P. 103079 - 103079

Published: Jan. 5, 2025

Language: Английский

---

Out-of-distribution generalization for segmentation of lymph node metastasis in breast cancer

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 7, 2025

Pathology provides the definitive diagnosis, and Artificial Intelligence (AI) tools are poised to improve accuracy, inter-rater agreement, turn-around time (TAT) of pathologists, leading improved quality care. A high value clinical application is grading Lymph Node Metastasis (LNM) which used for breast cancer staging guides treatment decisions. challenge implementing AI widely LNM classification domain shift, where Out-of-Distribution (OOD) data has a different distribution than In-Distribution (ID) train model, resulting in drop performance OOD data. This work proposes novel clustering sampling method automatically curate training datasets an unsupervised manner with aim improving model generalization abilities. To evaluate proposed models, we applied use Two One-sided Tests (TOST) method. examines whether on ID equivalent, serving as proxy generalization. We provide first evidence computing equivalence margins that data-dependent, reduces subjectivity. The framework shows ensembled models constructed from generalized across both tumor normal patches enhanced performance, achieving F1 score 0.81 unseen samples. Interactive viewing slide-level segmentations can be accessed PathcoreFlow™ through https://web.pathcore.com/folder/18555?s=QTJVHJuhrfe5 . Segmentation available at https://github.com/IAMLAB-Ryerson/OOD-Generalization-LNM

Language: Английский

---

Sustainable Development and Corporate Profitability: Data Mining Approach

Information Systems Frontiers, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 8, 2025

Language: Английский

---

Performance evaluation and comparative analysis of different machine learning algorithms in predicting postnatal care utilization: Evidence from the ethiopian demographic and health survey 2016

PLOS Digital Health, Journal Year: 2025, Volume and Issue: 4(1), P. e0000707 - e0000707

Published: Jan. 9, 2025

Postnatal care refers to the support provided mothers and their newborns immediately after childbirth during first six weeks of life, a period when most maternal neonatal deaths occur. In 30 countries studied, nearly 40 percent women did not receive postpartum check-up. This research aims evaluate compare effectiveness machine learning algorithms in predicting postnatal utilization Ethiopia identify key factors involved. The study employs techniques analyse secondary data from 2016 Ethiopian Demographic Health Survey. It predict predictors via Python software, applying fifteen machine-learning sample 7,193 women. Feature importance were used select top predictors. models’ was evaluated using sensitivity, specificity, F1 score, precision, accuracy, area under curve. Among four experiments, tenfold cross-validation with balancing Synthetic Minority Over-sampling Technique outperformed. From models, MLP Classifier (f1 score = 0.9548, AUC 0.99), Random Forest 0.9543, 0.98), Bagging 0.9498, 0.98) performed excellently, strong ability differentiate between classes. Region, residence, education, religion, wealth index, health insurance status, place delivery are identified as contributing that utilization. assessed models for forecasting usage. Ten-fold Oversampling produced best results, emphasizing significance addressing class imbalance healthcare datasets. approach enhances accuracy dependability predictive models. Key findings reveal regional socioeconomic influencing PNC utilization, which can guide targeted initiatives improve ultimately enhance child health.

Language: Английский

---

Prediction of urinary tract infection using machine learning methods: a study for finding the most-informative variables

BMC Medical Informatics and Decision Making, Journal Year: 2025, Volume and Issue: 25(1)

Published: Jan. 9, 2025

Urinary tract infection (UTI) is a frequent health-threatening condition. Early reliable diagnosis of UTI helps to prevent misuse or overuse antibiotics and hence antibiotic resistance. The gold standard for urine culture which time-consuming also an error prone method. In this regard, complementary methods are demanded. the recent decade, machine learning strategies that employ mathematical models on dataset extract most informative hidden information center interest prediction purposes. study, approaches were used finding important variables UTI. Several types machines including classical deep purpose. Eighteen selected features from test, blood demographic data found as features. Factors extracted such WBC, nitrite, leukocyte, clarity, color, blood, bilirubin, urobilinogen, factors test like mean platelet volume, lymphocyte, glucose, red cell distribution width, potassium, age, gender previous use determinative prediction. An ensemble combination XGBoost, decision tree, light gradient boosting with voting scheme obtained highest accuracy (AUC: 88.53 (0.25), accuracy: 85.64 (0.20)%), according Furthermore, results showed importance age This study highlighted potential suggested. approach 85.64%. Gender

Language: Английский

---