Abstract:
Background and Objective: [177Lu]Lu-DOTATATE-PRRT works by binding to SSRs, expressed in high quantities by NET cells, and delivering a cytotoxic effect through a beta-particle emitting radioactive isotope. Clinical studies have demonstrated the efficacy of PRRT in extending progression-free survival and improving quality of life for patients with advanced, inoperable NETs. However, the duration of PRRT response is highly variable, ranging from a few months to years. Furthermore, the assessment of treatment response in neuroendocrine tumours poses unique challenges and traditional criteria, such as RECIST and PERCIST, have not been capable of adequately evaluating the efficacy of PRRT. Radiomics, defined as the high-throughput extraction of textural features from imaging, could fulfil this need. The aim of this study was to develop survival models that predict tumour progression in patients with liver-metastasized NETs, using clinical and textural data from PET/CT scans obtained before and after PRRT. Additionally, we investigated whether radiomic features correlated with histological characteristics to explore the viability of a "virtual biopsy" through texture analysis. Methodology: We retrospectively queried our institutional database for patients with NET LM who underwent treatment with PRRT. Then, using the software LifeX, we segmented VOIs on the pre-treatment [68Ga]Ga-DOTA-peptide PET/CT. The same lesions were then segmented on the post treatment PET/CT and radiomics features were extracted. We then constructed different Cox proportional hazard models to explore the relationship between progression free survival time and one or more predictor variables. The Cox models we developed had increasing complexity: starting with only clinical variables, we progressively added radiomic variables. The models began with clinical variables, followed by clinical + pre-treatment radiomics, then clinical + pre-treatment radiomics + post-treatment radiomics, and finally clinical + pre-treatment radiomics + post-treatment radiomics + delta radiomics. Key Findings: -1. Impact of Radiomics on Prognostic Models: PET radiomics demonstrated a significant role in enhancing model performance. Specifically, integrating PET radiomics (particularly post-treatments scans) improved the C-Index in test runs and reduced the standard deviation in models relying only on clinical data. -2. Virtual Biopsy Potential: The study supports the concept of "virtual biopsy," where radiomics can approximate histological insights. These findings highlight the utility of radiomic features in predicting tumour histotypes and grading, potentially offering a non-invasive alternative to conventional biopsy when a specimen cannot be obtained. Conclusion: This dissertation provides evidence that PET radiomic analysis is a valuable tool for the non-invasive prognostication and stratification of liver NET metastases. The integration of PET-derived texture features into predictive models can improve clinical decision-making for PRRT candidates, offering insights into tumour biology and enhancing individualised treatment planning.
Neuroendocrine tumours