Towards Multimodal Diagnostics in Turner Syndrome: Low-Cost Facial, Voice, and Body Biomarkers

OBJECTIVES
The overall aim of this project is to advance in the consolidation and validation of low-cost, non-invasive diagnostic tools for Turner syndrome (TS). This project will pursue four complementary objectives:
i) to continue the analysis of multimodal data already collected in the context of the BeNeXT project from TS patients and matched controls, refining algorithms for reliable biomarker discovery;
ii) to clinically validate smartphone-based low-cost protocols for 3D facial biomarker acquisition in patient cohorts, extending methodologies previously validated in controls;
iii) to extract and interpret the first clinically meaningful conclusions from the analysis of TS patient voice samples, based on protocols already presented within the Iberian speech analysis community;
iv) to explore and develop deep learning models for estimating body biomarkers from 2D images, opening a new line of phenotyping with high translational potential.
Together, these objectives will strengthen the clinical value of BeNeXT by expanding its range of validated biomarkers, ensuring their applicability in real-world patient populations, and contributing to the development of robust translational tools for rare disease diagnosis.

WORKPLAN AND USE OF THE FUNDING
The 12-month workplan is structured into four work packages, designed to ensure a balanced combination of methodological development, clinical validation, and scientific dissemination.
WP1 (months 1–4). Processing and refinement of existing multimodal datasets. We will continue analyzing phenomic data already acquired from TS patients and controls, refining facial landmark detection, geometric morphometrics, and voice preprocessing pipelines. The goal is to generate reproducible, clinically relevant biomarker candidates.
WP2 (months 3–8). Validation of low-cost 3D facial biomarker acquisition in patients. New targeted acquisitions will be performed in clinical environments to validate smartphone-based protocols for 3D facial phenotyping. This step is critical to ensure the robustness of the BeNeXT pipeline when applied to patient cohorts.
WP3 (months 6–12). Voice analysis and preliminary conclusions. Voice samples from TS patients will be systematically analyzed using feature extraction and machine learning models to identify potential biomarkers. Preliminary conclusions will be drawn and disseminated to guide future larger-scale studies.
WP4 (months 5–12). Body biomarker estimation from 2D images using deep learning. We will initiate the development of convolutional and transformer-based deep learning architectures to estimate clinically meaningful body proportions and other anthropometric biomarkers from 2D images. This line will expand the BeNeXT framework to multimodal phenotyping that is scalable and cost-effective.
Use of funding. The funding will support pre-doctoral researchers contributing to data processing, algorithm development, and clinical validation, often within their final degree or master theses. Additional resources will be dedicated to dissemination at leading scientific conferences, covering fees, registration, and travel to ensure maximum visibility and impact.

EXPECTED RESULTS
By the end of the project, we expect to achieve the following key outcomes:
i) refined algorithms for 3D facial biomarker extraction validated in TS patients;
ii) preliminary voice biomarkers offering new insight into the phenotypic variability of TS;
iii) the first generation of deep learning models for estimating body biomarkers from 2D images, laying the foundation for scalable, low-cost body phenotyping;
iv) an expanded and technically validated BeNeXT framework integrating facial, voice, and body biomarkers.
Dissemination of results will target high-impact international conferences in medical imaging, speech technology, and computer vision (e.g., ISBI, Interspeech). These venues will provide rigorous peer validation while fostering collaborations across disciplines.
In addition, the project will support the training of young researchers through direct involvement in data analysis, model development, and dissemination, strengthening the link between academic innovation and clinical translation. Ultimately, this project will consolidate BeNeXT as a pioneering platform for low-cost, multimodal, and clinically validated biomarker discovery in Turner syndrome and rare diseases more broadly.