Advanced Quantification Pipeline Reveals New Spatial and Temporal Tumor Characteristics in Preclinical Multiple...

Area of Science:

• * Preclinical imaging
• * Cancer biology
• * Medical physics

Background:

• * Manual quantification of longitudinal small animal PET/CT for multiple myeloma (MM) is limited by annotator bias, excretion artifacts, and registration errors.
• * Developing a semi-automated pipeline can improve the characterization of tumor biology in MM.
• * Targeting regions of interest (ROIs) within the mouse skeleton, including complex areas like the pelvis, is crucial for accurate analysis.

Purpose of the Study:

• * To develop and validate a semi-automated PET/CT quantification pipeline for MM in preclinical models.
• * To achieve sub-organ spatial resolution for improved tumor characterization.
• * To analyze tumor distribution, burden, and bone involvement in a longitudinal study.

Main Methods:

• * An Attention U-Net model was trained to segment specific skeletal regions (spine, pelvis, sacrum, femurs) from CT slices.
• * A custom algorithm was used to mask physiological excretion spillover signals.
• * PCA-based projection mapped tumor distribution, and quantification metrics (SUVmean, SUVmax, HU) were calculated.

Main Results:

• * Tumor burden was preferentially localized to skeletal regions near joints.
• * Precise CT-based alignment (DICE = 0.966 ± 0.005) enabled detection of early disease progression.
• * Significant increases in tumor uptake (SUVmean) were observed across multiple skeletal sites by day 18.
• * Female mice exhibited greater bone loss near the hip joint at later stages, indicated by significant HUmean reductions.

Conclusions:

• * The developed pipeline enables reproducible and anatomically precise quantification of MM progression.
• * It accurately identifies region-specific trends, including joint tropism and sex-based differences.
• * This approach mitigates common challenges in manual analysis, enhancing the evaluation of tumor biology and treatment response in bone-involved cancer models.