RESUMO
Introduction: State of the art artificial intelligence (AI) models have the potential to become a "one-stop shop" to improve diagnosis and prognosis in several oncological settings. The external validation of AI models on independent cohorts is essential to evaluate their generalization ability, hence their potential utility in clinical practice. In this study we tested on a large, separate cohort a recently proposed state-of-the-art convolutional neural network for the automatic segmentation of intraprostatic cancer lesions on PSMA PET images. Methods: Eighty-five biopsy proven prostate cancer patients who underwent 68Ga PSMA PET for staging purposes were enrolled in this study. Images were acquired with either fully hybrid PET/MRI (N = 46) or PET/CT (N = 39); all participants showed at least one intraprostatic pathological finding on PET images that was independently segmented by two Nuclear Medicine physicians. The trained model was available at https://gitlab.com/dejankostyszyn/prostate-gtv-segmentation and data processing has been done in agreement with the reference work. Results: When compared to the manual contouring, the AI model yielded a median dice score = 0.74, therefore showing a moderately good performance. Results were robust to the modality used to acquire images (PET/CT or PET/MRI) and to the ground truth labels (no significant difference between the model's performance when compared to reader 1 or reader 2 manual contouring). Discussion: In conclusion, this AI model could be used to automatically segment intraprostatic cancer lesions for research purposes, as instance to define the volume of interest for radiomics or deep learning analysis. However, more robust performance is needed for the generation of AI-based decision support technologies to be proposed in clinical practice.
RESUMO
PURPOSE: The aim of this study is to investigate the role of [68Ga]Ga-PSMA-11 PET radiomics for the prediction of post-surgical International Society of Urological Pathology (PSISUP) grade in primary prostate cancer (PCa). METHODS: This retrospective study included 47 PCa patients who underwent [68Ga]Ga-PSMA-11 PET at IRCCS San Raffaele Scientific Institute before radical prostatectomy. The whole prostate was manually contoured on PET images and 103 image biomarker standardization initiative (IBSI)-compliant radiomic features (RFs) were extracted. Features were then selected using the minimum redundancy maximum relevance algorithm and a combination of the 4 most relevant RFs was used to train 12 radiomics machine learning models for the prediction of PSISUP grade: ISUP ≥ 4 vs ISUP < 4. Machine learning models were validated by means of fivefold repeated cross-validation, and two control models were generated to assess that our findings were not surrogates of spurious associations. Balanced accuracy (bACC) was collected for all generated models and compared with Kruskal-Wallis and Mann-Whitney tests. Sensitivity, specificity, and positive and negative predictive values were also reported to provide a complete overview of models' performance. The predictions of the best performing model were compared against ISUP grade at biopsy. RESULTS: ISUP grade at biopsy was upgraded in 9/47 patients after prostatectomy, resulting in a bACC = 85.9%, SN = 71.9%, SP = 100%, PPV = 100%, and NPV = 62.5%, while the best-performing radiomic model yielded a bACC = 87.6%, SN = 88.6%, SP = 86.7%, PPV = 94%, and NPV = 82.5%. All radiomic models trained with at least 2 RFs (GLSZM-Zone Entropy and Shape-Least Axis Length) outperformed the control models. Conversely, no significant differences were found for radiomic models trained with 2 or more RFs (Mann-Whitney p > 0.05). CONCLUSION: These findings support the role of [68Ga]Ga-PSMA-11 PET radiomics for the accurate and non-invasive prediction of PSISUP grade.
