Developing Analytical Pipelines for PET Imaging Patients with Alzheimer’s Disease

PET imaging is used in AD research and clinical evaluation to assess pathological protein accumulation. AD diagnosis relative to cognitively unimpaired individuals relies on standardized uptake value ratios and Centiloid scales. We introduce a data-science–driven, matrix-based framework for PET image interpretation.

PET datasets were obtained from the Data Image Archive (ida.loni.usc.edu/). 24 total PET datasets were analyzed, including 12 patients with AD and 12 CU controls matched for age and sex. The image-processing workflow was adapted from a previously published PET image visualization and computational analysis approach [1], with minor modifications. Image processing and analysis were performed using Fiji ImageJ (imagej.net) and 3D Slicer (slicer.org). Leveraging matrix-based computational approaches, raw PET images were treated as 8-bit grayscale data (intensity range: 0–255) and analyzed under three distinct processing paradigms: data-lossy, data-lossless, and data-enhanced.

In the data-lossy condition, Aβ signals resembled conventional SUVR and CL evaluations, showing minimal detectable brain signal, with higher intensity in AD than CU (AD > CU). In contrast, data-lossless processing revealed prominent Aβ-associated signals in both groups across multiple regions. Brain signal intensity was higher in AD than CU (AD > CU), whereas scalp and cervical lymph node (cLN) signal intensity was lower in AD (AD < CU). Data-enhanced visualization further resolved Aβ-associated signals into canal-like structures. These patterns occurred in both groups but showed greater apparent signal burden and structural congestion in AD (AD > CU). Aβ signal intensity in cLNs remained reduced in AD relative to CU (AD < CU), suggesting altered peripheral clearance pathways.

These findings suggest that PET datasets may encode spatial patterns of Aβ-associated tracer distribution not captured by conventional visualization. Aβ-associated signals within canal-like structures may be associated with brain clearance pathways, including lymphatic-related systems [2, 3]. These observations are exploratory; anatomical and physiological functions require further biological and quantitative validation.