Background:
Peptide receptor radionuclide therapy (PRRT) with [177Lu]-DOTATATE has demonstrated improvement in progression free survival for patients with metastatic pancreatic neuroendocrine tumors (pNETs), although treatment responses remain heterogeneous. Given the toxicity, cost and prolonged treatment course of PRRT, there is growing interest in imaging-derived markers capable of identifying early therapeutic response. Longitudinal assessment of somatostatin receptor expression (SSTR) using [68Ga]-DOTATATE PET/CT offers biologically relevant information, yet evidence linking PRRT response with conventional standardized uptake value (SUV) and volumetric features remains inconsistent. Topological data analysis (TDA) enables the quantification of spatially informed metrics descriptive of intratumoral structure and heterogeneity, with the potential to inform on therapy-induced effects that may precede measurable clinical responses. Hypothesis: Topological features from [68Ga]-DOTATATE PET identify PRRT-induced structural changes that are not reflected in SUV or volume, and expose response heterogeneity between lesions within the same patient.

Hypothesis/Objective:
Topological features from [68Ga]-DOTATATE PET identify PRRT-induced structural changes that are not reflected in SUV or volume, and expose response heterogeneity between lesions within the same patient.

Methods:
Seven metastatic lesions from one patient with pNET undergoing [177Lu]-DOTATATE PRRT were imaged with [68Ga]-DOTATATE PET/CT at three timepoints: before therapy (pre-PRRT), mid-way through treatment (mid-PRRT), and after completion (post-PRRT). Lesions were contoured using a semi-automated AI-based method and reviewed by an expert nuclear medicine physician. For each lesion at each timepoint we measured SUVmean, SUVmax, lesion volume, and three shape-based descriptors. The descriptors captured (1) the number of distinct hot-spot regions within the lesion, (2) ring-shaped uptake patterns surrounding cooler cores, and (3) enclosed three-dimensional low-uptake cavities. Per-lesion longitudinal trajectories of each measure were compared.

Results:
SUV and volume changed only modestly across treatment and did not distinguish lesions with clearly different underlying structural behavior. Shape-based analysis revealed heterogeneous response across the seven lesions of the same patient. One lesion showed progressive loss of internal structural complexity. The number of ring-shaped uptake patterns in this lesion fell from 60 to 15, consistent with structural consolidation typical of treatment response. A second lesion of similar pre-therapy size and uptake showed the opposite trajectory. The number of ring-shaped patterns in this lesion rose from 40 to 225, consistent with progressive internal restructuring despite minimal change in SUV. Post-PRRT emergence of enclosed low-uptake cavities was detected in three of seven lesions that had no such cavities at baseline. One lesion developed 33 new enclosed cavities post-PRRT where none existed before therapy. Two additional lesions showed post-PRRT fragmentation of previously contiguous hot-uptake regions into multiple separate foci. None of these patterns were detectable on SUVmean, SUVmax, or volume.

Conclusion:
Topological analysis of [68Ga]-DOTATATE PET/CT detected PRRT-induced structural changes in tumor uptake patterns that were not captured by standard SUV or volume measurements. Findings included therapy-induced cavity formation and divergent trajectories of internal restructuring between lesions in the same patient. As a single-patient case study, findings require replication. Evaluation in a larger cohort with clinical outcome data is ongoing. Post-PRRT cavity formation may serve as a candidate imaging biomarker of early therapy-induced necrosis during PRRT for metastatic pNETs.

Image/Figure:
Click to view full size
Click to view full size

Image/Figure Caption:
Longitudinal [⁶⁸Ga]-DOTATATE PET/CT and topological analysis of PRRT response in Patient 1. (A) Fused PET/CT images of a representative liver metastasis at pre-PRRT, mid-PRRT, and post-PRRT (top row); three-dimensional surface renderings of SUV within the same lesion at each timepoint (middle row); and box plots of lesion volume, SUVmean, and SUVmax across all seven lesions at each timepoint. Standard metrics changed only modestly across treatment. (B) Topological analysis of the same imaging data for two contrasting lesions in the same patient: Lesion 1 (left, progression-like pattern) and Lesion 2 (right, response-like pattern). For each lesion, three shape-based descriptors are plotted against SUV threshold: β₀ counts discrete hot-uptake regions; β₁ counts ring-shaped uptake patterns surrounding cooler cores; β₂ counts enclosed three-dimensional low-uptake cavities. Blue, orange, and green represent pre-PRRT, mid-PRRT, and post-PRRT respectively. Lesion 1 showed post-PRRT increase across all three descriptors, consistent with progressive internal restructuring. Lesion 2 showed the opposite trajectory. Persistence diagrams of β₁ features (bottom row) confirm this contrast: each point represents one loop-like feature in the lesion, with more points further from the diagonal indicating more prominent internal structure. Lesion 1 feature count rose across treatment (71, 134, 294 at pre-PRRT, mid-PRRT, and post-PRRT). Lesion 2 feature count fell across treatment (160, 37, 83).

Author
Carlos A. Gallegos
The University of Texas MD Anderson Cancer Center