Abstract
Low appendicular lean mass (ALM) is a biomarker that is associated with an increased risk of weakness, falls, functional decline, frailty, and adverse health outcomes, especially when combined with impaired strength or physical performance. ALM is usually determined using whole-body dual-energy X-ray absorptiometry (DXA) scans or by means of bioelectrical impedance analysis. However, both methods are uncommonly performed in routine clinical care. We therefore develop a deep learning model that employs hierarchical fusion mechanisms to use clinical data and DXA scans of hip and spine to estimate ALM. These scans are already collected during routine osteoporosis screening in millions of people worldwide. The proposed model uses two independent image encoders and a clinical data encoder to extract relevant features from each modality. These features are fused hierarchically using bidirectional feature-wise linear modulation. The combined framework outperforms single modalities by up to 47% in a cohort of more than 30,000 individuals. The clinical relevance is established by showing cross-sectional associations between predicted ALM and hand grip strength in more than 50,000 individuals. These results are the first step towards a scalable tool for ALM estimation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/AMAI_023.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/profile?id=%7EMarion_Mundt1
BibTex
@InProceedings{MunMar_Appendicular_MICCAISAT2026,
author = { Mundt, Marion AND Saleem, Afsah AND Gebre, Abadi K. AND Maqsood, Arooba AND Schousboe, John T. AND Leslie, William D. AND Harvey, Nicholas C. AND Gilani, Syed Zulqarnain AND Hu, Kun AND Suter, David AND Lewis, Joshua R. AND Sim, Marc},
title = { { Appendicular Lean Mass Estimation from Dual-View DXA Scans Using Hierarchical Fusion } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17273},
month = {pending},
page = {pending}
}
