Integrating Structured-light 3D Depth Sensing and Voice Analysis as a Multimodal AI Approach for Early Screening of Acromegaly: A Narrative Review and Future Perspectives
Year: 2025
Type: preprint
Abstract: Background. Acromegaly is a rare endocrine disorder characterized by the chronic hypersecretion of growth hormone (GH) and elevated insulin-like growth factor 1 (IGF-1), often diagnosed late due to the subtle early signs. While biochemical assays and pituitary MRI are accurate, they are unsuitable for population-wide screening. Structured-light (SL) 3D depth sensing and automated voice analysis can detect early craniofacial and phonatory changes. Objective. To evaluate the potential of a multimodal AI framework integrating SL and voice biomarkers for earlier, scalable acromegaly detection. Methods. A narrative review (2010-2025) of PubMed, Scopus, and IEEE Xplore identified peerreviewed studies on 2D/3D facial analysis, SL depth sensing, voice biomarkers, and multimodal AI. Diagnostic accuracy, technical performance, and methodological quality were synthesized using the QUADAS-2 framework. Results. SL achieved sub-millimeter accuracy (RMSE 0.78-0.92 mm), with sensitivity ranging from 85% to 92% and specificity ranging from 88% to 94%. Voice analysis detected a lowered fundamental frequency, increased jitter/shimmer, and a reduced harmonics-to-noise ratio, with an AUC of 0.78-0.86. In one multicenter study (n = 524), the AUC of 0.84 exceeded the endocrinologists' 0.69. No study has tested both modalities in the same cohort. Discussion. SL captures stable morphological features, while voice analysis reflects earlier functional changes; integration could enhance the early detection of these changes. Translation to practice requires diverse, representative datasets, standardized acquisition protocols, and hardware-agnostic compatibility. Longitudinal monitoring could boost sensitivity and reduce false negatives. Economic modeling supports the cost-effectiveness of a tiered pathway. Conclusion. If validated in large, multicenter trials, this approach may reduce diagnostic delays in acromegaly and serve as a blueprint for multimodal AI in other rare, high-impact diseases.
Author: M
Language: English
Subfield: Endocrinology, Diabetes and Metabolism
Field: Medicine
Domain: Health Sciences
Open Access status: gold
OpenAlex ID: W4413638626