The study investigates why machine‑learning models for tuberculosis screening based on cough acoustics fail to generalize across datasets. Classical ML and deep‑learning classifiers achieved moderate performance within their own datasets (ROC‑AUC up to 0.755) but performed poorly on external data, often below 0.6. The authors found that audio features were more influenced by recording device and dataset than by TB status, and that device‑diverse training improved transfer while device mismatch degraded it. A clinical‑variable baseline showed more consistent generalization, suggesting acquisition‑specific variability is a stronger driver of poor generalizability than population shift.
arXiv:2608.21427v1 Announce Type: new Abstract: Tuberculosis (TB) is one of the most common and dangerous bacterial ailments. Every year, it causes a large number of deaths worldwide. Although many d...
arXiv:2606. 15436v1 Announce Type: cross Abstract: Respiratory acoustic foundation models (FMs) excel at cough classification, yet their ability to predict continuous health quantities from cough audio remains largely unexplored, despite the clinical value of passive age, BMI, and disease probability estimation in settings where physical measurements are unavailable.
The study evaluates machine learning models—logistic regression, multilayer perceptrons, and convolutional neural networks—on clinical and demographic data from Uganda and South Africa to screen for tuberculosis. Feature selection improves AUROC by 2–7% across models, with logistic regression achieving 0.80–0.84 AUROC on held‑out data. While deeper networks show inconsistent performance on external cohorts, logistic regression remains within 1–2% of the best AUROC and narrowly misses WHO sensitivity targets.
arXiv:2610.15702v1 Announce Type: cross Abstract: Respiratory auscultation provides a low-cost means of assessing pulmonary function, yet transient crackles and sustained wheezes can be difficult to...
The study compares classical and deep learning classifiers for a five‑class respiratory‑sound dataset of 1,211 clips. An SVM using 30 summary features (MFCC, ZCR, RMS) achieves macro recall of 0.78, while a 1.13M‑parameter convolutional‑recurrent network reaches 0.75–0.77; no clear advantage is found between model families. The largest performance gain comes from standardizing the SVM’s features, which boosts macro recall by 27 points, far exceeding any architectural difference.
arXiv:2606. 02998v1 Announce Type: new Abstract: Automated cough analysis offers a path to low-cost respiratory screening, but most existing work stops at binary COVID-19 detection.
The paper introduces BTS-CAFE, a federated domain generalization framework for respiratory sound classification that addresses stethoscope-induced shortcuts. It combines causality-inspired device-style interventions, counterfactual metadata augmentation, and gradient alignment to reduce style–content entanglement and promote device-invariant decision boundaries. Experiments on ICBHI and SPRSound datasets show a 3.69‑point improvement in out-of-distribution performance over the baseline and outperform conventional data augmentation and federated learning methods.
The paper introduces LLM4CKD, a framework that uses large language models (LLMs) for early chronic kidney disease (CKD) screening without task‑specific training. By employing clinically selected tabular features and structured prompt templates, the authors evaluate LLMs in zero‑shot and few‑shot settings against traditional machine learning, deep learning, and tabular foundation models. Results show that LLMs can match or outperform conventional methods in low‑data scenarios, though their performance varies with model choice and input complexity, highlighting a trade‑off between data efficiency and stability.
Deep-learning models can achieve strong chest X-ray (CXR) classification performance without establishing whether their predictions predominantly rely on pulmonary image content. This study evaluates...
arXiv:2609.27222v1 Announce Type: cross Abstract: Pneumonia is difficult to diagnose in older long-term care residents; multimorbidity and atypical presentations obscure signs, motivating operational...
arXiv:2608.21583v1 Announce Type: new Abstract: Oral cancer is a leading cause of mortality in low-to-middle-income countries, where a shortage of specialists delays diagnosis. While point-of-care sc...
The study evaluates the robustness of medical vision‑language models for tuberculosis screening on chest X‑rays by testing them across multiple datasets, prompts, and evaluation settings. Three specialized models (BioMedCLIP, CheXficient, MedSigLIP) and a general OpenCLIP model were audited on 12,200 images, producing 244,000 model–image–prompt scores. Results show that no model consistently outperforms others across all cohorts and reliability criteria, with prompt changes and control group composition significantly affecting AUROC, and that high training‑set performance does not reliably transfer to external cohorts.