Two-Stream Convolutional Neural Networks for Breathing Pattern Classification: Real-Time Monitoring of Respiratory Disease Patients

Bioengineering, Год журнала: 2024, Номер 11(7), С. 709 - 709

Опубликована: Июль 12, 2024

A two-stream convolutional neural network (TCNN) for breathing pattern classification has been devised the continuous monitoring of patients with infectious respiratory diseases. The TCNN consists a (CNN)-based autoencoder and classifier. encoder generates deep compressed feature maps, which contain most important information constituting data. These maps are concatenated generated by classifier to classify patterns. TCNN, single-stream CNN (SCNN), state-of-the-art models were applied four patterns: normal, slow, rapid, breath holding. input data consisted chest tissue hemodynamic responses measured using wearable near-infrared spectroscopy device on 14 healthy adult participants. Among evaluated, random forest had lowest accuracy at 88.49%, while achieved highest 94.63%. In addition, proposed performed 2.6% better in terms than an SCNN (without autoencoder). Moreover, mitigates issue declining learning performance increasing depth, as observed model. results prove robustness classifying patterns despite significantly smaller number parameters computations compared models.

Язык: Английский

---

Implementing Explainable Machine Learning Models for Practical Prediction of Early Neonatal Hypoglycemia

Diagnostics, Год журнала: 2024, Номер 14(14), С. 1571 - 1571

Опубликована: Июль 19, 2024

Hypoglycemia is a common metabolic disorder that occurs in the neonatal period. Early identification of neonates at risk developing hypoglycemia can optimize therapeutic strategies care. This study aims to develop machine learning model and implement predictive application assist clinicians accurately predicting within four hours after birth. Our retrospective analyzed data from born ≥35 weeks gestational age admitted well-baby nursery between 1 January 2011 31 August 2021. We collected electronic medical records 2687 tertiary center Southern Taiwan. Using 12 clinically relevant features, we evaluated nine approaches build models. selected models with highest area under receiver operating characteristic curve (AUC) for integration into our hospital information system (HIS). The top three AUC values early prediction were 0.739 Stacking, 0.732 Random Forest Voting. considered best because it has relatively high shows no significant overfitting (accuracy 0.658, sensitivity 0.682, specificity 0.649, F1 score 0.517 precision 0.417). was incorporated web-based integrated system. Shapley Additive Explanation (SHAP) indicated mode delivery, age, multiparity, respiratory distress, birth weight < 2500 gm as five predictors hypoglycemia. implementation provides an effective tool assists identifying at-risk hypoglycemia, thereby allowing timely interventions treatments.

Язык: Английский

---

Online performance prediction using the fusion model of LightGBM and TabNet for large laser facilities

International Journal of Data Science and Analytics, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 20, 2024

Язык: Английский

---

Simulation of the Possible Routes of Acinetobacter spp. Transmission in the Intensive Care Units: An Agent-Based Computational Study

Iranian Journal of Medical Microbiology, Год журнала: 2024, Номер 18(5), С. 287 - 300

Опубликована: Ноя. 30, 2024

Язык: Английский

---