P-113: An artificial intelligence-based 3D-segmentation assessment of sarcopenia utilizing whole-body computed tomography in patients with multiple myeloma

Introduction: Sarcopenia is characterized by loss of muscle mass and strength, resulting in functional limitations and an increased risk of falls, injuries and fractures. Aim of this study was to obtain detailed information on skeletal muscle changes in patients with multiple myeloma (MM) during treatment. Therefore, an artificial intelligence-based 3D-segmentation method was established to assess entire muscle groups.

Methods: Patients diagnosed with MM (n=51) who had undergone both initial (T1) and follow-up (T2) whole-body low-dose computed tomography acquisition prior to induction therapy and post autologous hematopoietic stem cell transplantation were examined retrospectively. Muscle volume (MV) and intramuscular adipose tissue volume (IMAT) of the autochthonous back muscles, the iliopsoas muscle and the gluteal muscles were evaluated. The tools TotalSegmentator and AutoMatiCA were combined with a separate python script. Subsequently, an artificial intelligence-network was trained in order to obtain a fully automated 3D-segmentation assessment.

Results: Patients‘ median age was 58y, 38 were male and follow-up CT-scans were performed after a mean of 11.8 months. Changes in MV and IMAT correlated significantly with body mass index (BMI, p < 0.0001). Patients (n=28) with a decrease of BMI (mean -2.2 kg/m²) during therapy lost MV (T1: 3419cm³, IQR 3176-4000cm³ vs. T2: 3226cm³, IQR 3014-3662cm³, p < 0.0001) whereas patients (n=20) with an increased BMI (mean +1.4 kg/m²) showed an increase of IMAT (T1: 122cm³, IQR 96.8-202.8cm³ vs. T2: 145.5cm³, IQR 115-248cm³, p=0.0002). Loss of MV varied between different muscle groups and was most prominent in the iliopsoas muscle(-9.8%) > gluteus maximus(-9.1%) > gluteus medius(-5.8%) > autochthonous back muscles(-4.3%) > gluteus minimus (-1.5%). Increase of IMAT in patients who gained weight was similar between muscle groups.

Conclusions: The artificial intelligence-based 3D-segmentation process is a reliable and time-saving method to acquire in depth information on sarcopenia in MM patients. Loss of MV and increase of IMAT were reliably detectable and associated with changes in BMI. Loss of MV was highest in muscles with more type 2 muscle fibers (fast-twitch, high energy) whereas muscles with predominantely type 1 fibers (slow-twitch, postural control) were less affected. In order to acquire additional information on physical function and training effects, we are currently using this 3D-segmentation method in a randomized controlled trial on an exercise intervention during induction therapy.