Fully Automated Image-Based Multiplexing of Serial PET/CT Imaging for Facilitating Comprehensive Disease Phenotyping

Area of Science:

• Medical Imaging
• Radiochemistry
• Computational Biology

Background:

• Combined PET/CT imaging offers insights into anatomic and molecular processes.
• Traditional single-tracer approaches limit comprehensive disease phenotyping.
• Multidimensional analysis requires integrating data from serial PET/CT scans.

Purpose of the Study:

• To develop an open-source postprocessing tool, the PET Unified Multitracer Alignment (PUMA) framework.
• To enable multiplexing of serial PET/CT scans for comprehensive voxelwise tissue characterization.
• To enhance multidimensional disease phenotyping using multitracer PET/CT imaging.

Main Methods:

• PUMA utilizes AI-based CT segmentation for multilabel maps of 24 body regions.
• A two-step registration process (affine followed by symmetric diffeomorphic) is employed.
• Tracer images are normalized and assigned to RGB channels for simultaneous visualization of up to three tracers.

Main Results:

• Deformable registration achieved superior alignment (Dice > 0.90 in 60% of organs) with minimal intensity differences (<3%).
• PUMA preserved quantitative PET data, showing minimal SUVmean differences in tumors (1.6% ± 0.9%).
• The framework demonstrated robust spatial multiplexing capabilities across multiple centers and vendors.

Conclusions:

• PUMA offers a vendor-independent solution for post-acquisition multiplexing of serial PET/CT images.
• It integrates complementary tracer data voxelwise into a composite image without altering clinical protocols.
• This approach enhances multidimensional disease phenotyping and supports improved diagnostic and therapeutic decisions.