Non-Invasive Computational Diagnostics for Laryngeal Pathology: Voice-Based and Multimodal Learning Perspectives

Timely and precise identification of laryngeal pathologies plays a critical role in preventing cancer progression and optimizing treatment effectiveness. Contemporary developments in machine learning and computational diagnostics have facilitated the creation of both imaging-dependent invasive techniques and voice-analysis-based non-invasive solutions for detecting laryngeal and throat malignancies. This work offers a comparative evaluation of these computational methodologies, examining their diagnostic capabilities, interpretive transparency, and practical clinical utility. Evidence from contemporary research demonstrates that deep learning architectures utilizing invasive imaging modalities, especially convolutional neural networks trained on endoscopic or tissue biopsy data, deliver exceptional diagnostic precision, frequently surpassing 94% accuracy. In contrast, non-invasive acoustic analysis methods that employ voice features including MFCC, PLP, jitter, and shimmer represent accessible and economical alternatives, yielding accuracy rates ranging from $82 \%$ to $\mathbf{9 0 \%}$. Although their precision is somewhat reduced, these approaches offer considerable advantages in simplicity and patient acceptance, rendering them particularly valuable for preliminary screening programs and remote medical consultation services. This research emphasizes the expanding significance of integrated multimodal computational systems that synthesize acoustic signals, medical imaging, and clinical parameters to strengthen diagnostic reliability and robustness. The study addresses continuing obstacles, such as unbalanced training datasets, model performance across varied recording environments, and the explainability of sophisticated learning models. The findings establish that voice-based computational diagnostics constitute a significant advancement toward interpretable, non-invasive, and patient-focused technologies in laryngeal healthcare.