Comparison of 18F-Choline and 18F-PSMA Performance On Detecting Malignant Lesions Using Fuzzy C-Means Segmentation On PET-CT Images of Patients with Recurrent Prostate Cancer
M Katsikakis1, I Gatos1, N Papathanasiou2, S Tsantis1, D Apostolopoulos2, K Papadimitropoulos2, M Spiliotopoulou2, E Liatsikos3, D Mihailidis4, G C Kagadis1*, (1) Department of Medical Physics, University of Patras, Rion, GR 26504, GR, (2) Department of Nuclear Medicine, University of Patras, Rion, GR 26504, GR, (3) Department of Urology, University of Patras, Rion, GR 26504, GR, (4) University of Pennsylvania, Wynnewood, PA, USA
Presentations
PO-GePV-I-78 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: The objective of the present study is the comparison of the sensitivity in detecting malignancies between two radioactive tracers, 18F-Choline and 18F-PSMA. Comparison was based on patients’ image processing and analysis findings who underwent PET-CT imaging with both formulations. Direct juxtaposition of their differences in concentration and absorption of suspicious areas was performed.
Methods: The clinical dataset included 25 male patients (mean age: 70.68±6.6 SD) evaluated for biochemical recurrence of prostate cancer with increased PSA levels >0.20 ng/ml, after initial radical treatment. Every patient was prospectively evaluated with both 18F-Choline and 18F-PSMA PET/CT within a week according to approved imaging protocols. For each patient, the attenuation corrected (AC) PET-CT axial series for both radiopharmaceuticals were extracted. A semi-automatic segmentation algorithm was employed for each set by firstly involving Minimum Intensity Projection (MinIP) to a single coronal image representing total uptake of the tracer throughout the patient’s anatomy. Consequently, manual cropping of areas of the resulting image containing suspicious findings by an expert Nuclear Medicine Physician was performed, followed by Fuzzy C-Means (FCM) automatic clustering for suspicious findings’ segmentation, and mean intensity calculation. A total of 95 suspicious findings originating from both tracers, were extracted using the aforementioned procedure. An ROC analysis was performed on mean segmented ROIs’ intensity measurements for each method’s performance on ROI malignancy detection using the expert’s indication (benign=0, malignant=1) as ‘Gold Standard’ after full diagnosis of each patient and having followed their clinical progress.
Results: 18F-PSMA and 18F-Choline accuracies in detecting malignancies were 77.35% and 66.07% using an intensity threshold of ≤ 210.40 and ≥ 43.14 respectively.
Conclusion: The results show that 18F-PSMA is more accurate in diagnosing and differentiating suspicious findings as malignant or benign and demonstrates lower uptake in healthy tissues than 18F-Choline.
Funding Support, Disclosures, and Conflict of Interest: This study is part of a prospective, multicentre trial funded by "Biokosmos" Company (radiopharmaceutical production), Lavrion, Greece, and organized by "Pharmassist" Company.
Keywords
Segmentation, Image Processing, Image Analysis
Taxonomy
IM- PET : Quantitative imaging
Contact Email