Automated 68Ga-PSMA-PET Total Metabolic Tumor Volume (TMTV) Extraction for Predicting PSA Change in 177Lu-PSMA Therapy
Résumé
Background: Advanced prostate cancer remains a significant challenge in oncology. Radiopharmaceutical therapy with 177Lu-PSMA has emerged as a promising treatment, with 68Ga-PSMA PET/CT imaging playing a crucial role in evaluating disease extent and treatment response. This study aims to assess the correlation between PET/CT imaging features and serum PSA levels to refine patient eligibility and predict therapy outcomes.
Methods: Data from 78 male patients undergoing 177Lu-PSMA therapy (1–6 cycles of 7.4 GBq) were analyzed. Imaging included 68Ga-PSMA PET/CT and SPECT post-treatment. Automated segmentation using "TotalSegmentator" delineated lesions, excluding physiological uptake in organs-at-risk. Total Metabolic Tumor Volume (TMTV) was calculated by subtracting organ signals and applying patient-specific thresholds. PSA nadir, defined as the lowest PSA value during treatment, was used to classify responders (≥50% reduction in PSA). Mann-Whitney U-tests and a multilayer perceptron (MLP) model assessed correlations and classification performance.
Results:
A significant difference in mean standardized uptake value (SUV) of TMTV was observed between responders and non-responders (p = 0.001).
Classification using MLP achieved an F1-score of 0.71 for a balanced dataset and 0.48 for an unbalanced dataset.
A ROC curve derived from a balanced dataset with a universal SUV threshold of 2.57 yielded an F1-score of 0.75.
Conclusion: The study highlights a weak but statistically significant correlation between TMTV SUV mean and PSA response. The results emphasize the need for larger, curated datasets and refined TMTV computation for better prediction accuracy.
Origine | Fichiers produits par l'(les) auteur(s) |
---|---|
licence |