AI-based body composition analysis of CT data has the potential to predict disease course in patients with multiple myeloma.
Authors
Affiliations (10)
Affiliations (10)
- Institute of Interventional Radiology, University Hospital Schleswig-Holstein, Lübeck, Germany. [email protected].
- Fraunhofer Research Institution for Individualized and Cell-Based Medical Engineering, Fraunhofer IMTE, Lübeck, Germany. [email protected].
- Institute of Interventional Radiology, University Hospital Schleswig-Holstein, Lübeck, Germany.
- Institute of Radiology and Nuclear Medicine, University Hospital Schleswig-Holstein, Lübeck, Germany.
- Clinic of Rheumatology and Clinical Immunology, University Hospital Schleswig-Holstein, Lübeck, Germany.
- Department of Hematology and Oncology, University Cancer Center Schleswig-Holstein, University Hospital Schleswig-Holstein, University of Lübeck, Lübeck, Germany.
- Institute of Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany.
- Institute for Artificial Intelligence in Medicine, University Hospital Essen, Essen, Germany.
- Department of Hematology and Oncology, University Cancer Center Schleswig-Holstein, University Hospital Schleswig-Holstein, University of Lübeck, Lübeck, Germany. [email protected].
- Department of Hematology and Oncology, University Cancer Center Schleswig-Holstein, University Hospital Schleswig-Holstein, University of Lübeck, Lübeck, Germany. [email protected].
Abstract
The aim of this study was to evaluate the benefit of a volumetric AI-based body composition analysis (BCA) algorithm in multiple myeloma (MM). Therefore, a retrospective monocentric cohort of 91 MM patients was analyzed. The BCA algorithm, powered by a convolutional neural network, quantified tissue compartments and bone density based on routine CT scans. Correlations between BCA data and demographic/clinical parameters were investigated. BCA-endotypes were identified and survival rates were compared between BCA-derived patient clusters. Patients with high-risk cytogenetics exhibited elevated cardiac marker index values. Across Revised-International Staging System (R-ISS) categories, BCA parameters did not show significant differences. However, both subcutaneous and total adipose tissue volumes were significantly lower in patients with progressive disease or death during follow-up compared to patients without progression. Cluster analysis revealed two distinct BCA-endotypes, with one group displaying significantly better survival. Furthermore, a combined model composed of clinical parameters and BCA data demonstrated a higher predictive capability for disease progression compared to models based solely on high-risk cytogenetics or R-ISS. These findings underscore the potential of BCA to improve patient stratification and refining prognostic models in MM.