RESEARCHERS have developed an AI-based pipeline capable of extracting body composition metrics from whole-body MRI in patients with multiple myeloma, revealing that treatment-related changes in muscle and fat are linked to survival outcomes.
Patients with multiple myeloma are frequently exposed to prolonged, multi-line, multi-drug treatment that can substantially affect overall health, physical reserve, and physiological resilience. Organs captured incidentally by whole-body scans can be opportunistically interrogated to derive quantitative measures of body composition. This is the first reported development of an AI-based pipeline for automated, opportunistic phenotyping of non-diseased tissue from whole-body MRI in multiple myeloma.
The exploratory, single-centre study used a deep learning pipeline, originally trained on UK Biobank population imaging, that was retrained to extract image-derived phenotypes from routinely acquired clinical whole-body MRI in patients with multiple myeloma. The pipeline quantified abdominal skeletal muscle and abdominal subcutaneous and visceral adipose tissue at baseline and longitudinally through treatment, with outcomes related to progression free survival using models adjusted for conventional clinical metrics.
Significant longitudinal changes were observed throughout treatment (p<0.001), characterised by a decrease in abdominal skeletal muscle alongside transient increases in subcutaneous and visceral adipose tissue. Greater baseline reserves of abdominal skeletal muscle (HR 0.60, 95% CI 0.40 to 0.89) and abdominal subcutaneous adipose tissue (HR 0.67, 95% CI 0.46 to 0.98) were associated with better progression free survival, while increases in visceral adipose tissue over time were associated with inferior progression free survival (HR 2.89, 95% CI 1.65 to 5.09). The model achieved a concordance index of 0.725 for risk stratification.
Implications for Risk Stratification and Care
These findings have extended the clinical utility of whole-body MRI beyond lesion detection, supporting opportunistic body composition phenotyping as a scalable biomarker for risk stratification, treatment planning, and survivorship care in multiple myeloma. As this remains a single-centre, exploratory study with modest sample sizes at later time points, external validation in multicentre cohorts is needed before these image-derived phenotypes can inform routine clinical decisions.
Reference
Basty N et al. AI derived whole-body MRI metrics in multiple myeloma patients reveal unique insights into body composition and outcomes. Blood Advances. 2026;DOI:10.1182/bloodadvances.2026020852.
Featured image: LIGHTFIELD STUDIOS on Adobe Stock
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
