Blast cell segmentation and leukemia classification using hybrid Deep Kronecker WideResNet using blood smear images.

Predicting emergency mortality risk in traumatic brain injury: comparative analysis of machine learning and large language model GPT-5.

Tunable triazole-based cholera toxin inhibitors: A QSAR-guided design and evaluation approach.

MethylMSI: Prediction of microsatellite instability based on DNA methylation profile and SVM model.

FTIR spectroscopy combined with machine learning reveals molecular signatures distinguishing three phenotypes of endometriosis.

Multi-objective decision model for wastewater treatment technology selection based on machine learning.

Classification of DNA secondary structures by combining multiple spectral techniques with machine learning.

Machine learning-based prediction of three-year mortality in elderly inpatients with coronary artery disease combined with heart failure.

Machine learning (ML) is increasingly being used in food science due to its ability to extract insights from large datasets. However, the advantages of ML over traditional mechanistic knowledge-based models remain unclear, especially under the limited data conditions often encountered in food bioprocesses. This study aims to address this gap by critically evaluating supervised ML techniques-specifically decision trees, support vector machines, and neural networks-in comparison to a knowledge-based model (KB), using wine fermentation as a practical, experimental example. We evaluated these approaches in three tasks. Tasks 1 and 2 use time-series fermentation data to (1) classify industrial yeast strains based on their metabolite profiles and (2) predict fermentation dynamics. Task 3 focuses on creating a fast surrogate model using ML techniques applied to synthetic data generated by a mechanistic model. For yeast strain classification, we achieved our highest test accuracy of 74% when utilizing all available metabolite data. In predicting fermentation dynamics, the KB model outperformed the ML models, achieving an average normalized root mean squared error of approximately 6%. The ML models, when additional data was incorporated, had a prediction error of around 7.6%. Lastly, a deep learning surrogate model trained solely on synthetic, mechanistic data demonstrated very low errors (around 0.6%) on test sets, compared to the KB model, while also reducing simulation time by a factor of 30. Our findings highlight the significance of experimental design: although ML models perform well when trained on large and diverse datasets, they often struggle with limited data or when predicting outcomes beyond the conditions observed during training. In contrast, mechanistic models show better generalization and biological interpretability. The complementary nature of both approaches suggests that combining them can lead to more robust, data-informed design and control in complex fermentation systems. Leveraging these complementary strengths, we developed and validated a hybrid model that integrates knowledge-based predictions with a residual neural network to correct systematic errors, reducing overall NRMSE from 6% to 5% and improving prediction for most key compounds.

Smart three-in-one detection platform: Chemical sensing, image recognition, and machine learning for rapid identification of tetracycline antibiotics.

Delayed chemotherapy-induced nausea and vomiting (CINV) in pediatric oncology patients is currently under-recognized. This study aims to develop, validate, and visualize a machine learning-based model to predict delayed CINV risk in children. This prospective cohort study was conducted from November 2021 to December 2022 at a tertiary hospital in southern China. Pediatric delayed CINV data were collected via an electronic diary using the Pediatric Nausea Assessment Tool (PeNAT) and National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI-CTCAE) (v4.03), with PeNAT ≥3 or CTCAE grade ≥2 as the primary outcomes. Seven machine learning models, including random forest, support vector machine, and artificial neural network (ANN), were developed and validated using 29 sociodemographic and clinical features. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) and other metrics. Shapley's Additive Explanations (SHAP) enhanced interpretability, and the models were integrated into a web-based calculator for visualization. Overall, 399 pediatric patients (60.4% male; aged 4-18 years) were included. The AUC of the seven models ranged from 0.782 to 0.815, with the ANN model performing best (AUC 0.815; 95% CI, 0.695-0.903). The ANN model's global SHAP plot revealed that the most influential features were acute CINV, days of chemotherapy, age, number of recreational activities, expectancy of CINV, and control effectiveness of CINV. The ANN model was then deployed as a web-based risk calculator for pediatric delayed CINV. The ANN model demonstrated good performance in identifying children at high risk of delayed CINV. Our web-based calculator provides a reliable tool for clinical staff to support targeted CINV management.

Tracing the geographical origin of tiger nut (Cyperus esculentus L) in China based on stable isotopes and mineral elements combined with multi-modal recognition.

Machine learning classification of PAH exposure in the Antarctic limpet Nacella concinna.

Interpretable machine learning model for predicting refeeding syndrome after colorectal cancer surgery.

Establishment of a quantitative GC-MS method for acrylamide detection and in situ kinetic study of acrylamide formation in fried potato slices.

Specific diagnostic model for bacterial pneumonia constructed by combining multiple omics and multi machine learning models.

Interpretable machine learning-based prediction of liver metastasis risk in elderly patients with small cell lung Cancer: A study based on the SEER database and external validation in a Chinese cohort.

High-throughput identification of geographical origins of rubies using hyperspectral visible and fluorescence spectroscopy.

Aircraft pilots can be faced with a high mental workload (MW) combined with moderate hypoxia and sleep restriction. We aimed to assess the cross-validation of a machine learning-based MW predictive model under hypoxia and/or sleep restriction. Secondly, we developed a robust predictive model using multimodal physiological parameters to improve the validity across different physiological conditions. Seventeen healthy participants were randomly exposed to three 12-minute periods of increased MW (low, medium, and high) in a 4-condition crossover design: sleep restriction (SR, <3 h Total Sleep Time, TST) vs. habitual sleep (HS, >6 h TST), hypoxia (HY, 2 h, FIO2=13.6%, ∼3500 m) vs. normoxia (NO, FIO2=21%). MW levels were designed using the Multi-Attribute Test Battery (MATB)-II with an additional auditory Oddball-like task. Six machine learning classifiers were compared. Features selection (from EEG, ECG, respiratory and eye tracking sensors) was performed using backward Recursive Feature Elimination (RFE). The best models for 1-minute MW levels classification on HSNO were K-Nearest Neighbors (KNN, F1 score = 80.3 ± 8.9%), Support Vector Machine (SVM, 77.8 ± 10.3%) and Random Forest (RF, 75.7 ± 9.1%). Exposure to sleep restriction and/or hypoxia decreased models' performance (F1 <35%). KNN and RF models, in particular those including EEG and eye tracking, trained on All-Conditions performed well across conditions (F1 scores = 77.4 ± 7.8% and 70.7 ± 10.2%). Our results highlight the need for training MW models under different physiological constraints and using multimodal datasets to improve robustness. (NCT05563688).

Development and validation of interpretable machine learning models for dynamic prediction of prognosis in acute pancreatitis complicated by acute kidney injury: A multicenter study.