Chronic Obstructive Pulmonary Disease (COPD) is a leading cause of morbidity and mortality worldwide, with high underdiagnosis rates due to limitations in current diagnostic methods such as spirometry. This doctoral thesis explores the potential of voice as a digital biomarker to support the assessment of COPD, guided by the principles of Applied Health Technology (AHT), which emphasizes interdisciplinary collaboration and real-world applicability.

The research includes four interconnected studies. Study I presents a systematic literature review of machine learning (ML) applications for voice-affecting disorders, identifying COPD as underrepresented in current research. Study II addresses this gap by collecting a new dataset of vowel [a:] recordings from Swedish-speaking COPD patients and healthy controls once a week in self-determined quiet settings. Voice features, including baseline acoustic (BLA) parameters and Mel-Frequency Cepstral Coefficients (MFCCs), were extracted and used to train three ML classifiers: CatBoost (CB), Random Forest (RF), and Support Vector Machine (SVM). CB demonstrated the highest test accuracy at 78%. 

Study III investigates the effects of signal segmentation on model performance and shows that certain temporal segments of voice recordings contain more informative patterns, enhancing classification outcomes by increasing accuracy to 85%. Study IV applies statistical and practical significance tests to compare voice features between COPD and healthy groups. A total of 34 features, including shimmer measures and higher-order MFCC derivatives, were found to meaningfully differentiate the groups. 

This thesis reframes the human voice as a source of clinically relevant data, demonstrating how it can be digitized, analyzed, and interpreted using ML to aid COPD assessment. The results indicate that voice-based analysis can provide an accessible, non-invasive, and scalable complement to existing diagnostic tools. By integrating technical, clinical, and ethical perspectives, the thesis contributes new knowledge and practical methodologies that align with AHT's goal of creating value-driven, user-centered healthcare solutions. The findings support future development of mobile and remote voice-based screening tools for COPD and other conditions.