REPRODUCIBILITY OF RADIOMICS FEATURES IN 68GAPSMA-11 PET/CT IMAGES OF PATIENTS WITH METASTATIC PROSTATE CANCER FOR A PERSONALIZED TREATMENT APPROACH

Abstract

Background: Prostate-specific membrane antigen= Positron Emission tomography (PSMA PET) images' manual interpretation can lead to a considerable number of missed metastatic Prostate cancer lesions. Therefore, Radiomics has been utilized as a novel high-potential method. However, its features lack sufficient evidence regarding their robustness and reproducibility. Consequently, we opted to assess such features in cases with metastatic Prostate cancer.

Materials and Methods: Our study was carried out in a tertiary referral center during 2021 and 2022 on prostate cancer cases undergoing 68 Ga-PSMA-PET/CT imaging. Initially, the PET/CT scan findings were analyzed using Python's PyRadiomics tool to extract first-order and tissue-related characteristics. Then, the segmentation of the said features was manually carried out and classified into their respective categories. Their repeatability was ultimately determined by measuring the Intraclass Correlation Coefficients (ICCs).

Results: A total of 150 PSMA-PET/CT images were investigated, leading to the extraction of 101 features, among which 1, 3, and 10 had excellent, good, and moderate repeatability, respectively. Out of all the ICC values, only three were statistically significant. These values were for Low Gray-Level Run Emphasis (LGLRE; ICC = 0.799, p = 0.047) and Long-Run Low Gray-Level Emphasis (LRLGLE; ICC = 0.801, p = 0.045) in the Gray-level run length matrix as well as Low Gray-Level Emphasis (LGLE; ICC = 0.906, p = 0.003) in the Gray-level dependence matrix.

Conclusion: Several PSMA-PET/CT-derived radiomics features have significant metastasis-predicting values in metastatic Prostate cancer. However, future studies must assess the agreement between these features and clinical and histological parameters.

DOI: 10.46121/pspc.52.4.6