Dataset Information

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative ⁶⁸Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

ABSTRACT: High-resolution intraoperative PET/CT specimen imaging, coupled with prostate-specific membrane antigen (PSMA) molecular targeting, holds great potential for the rapid ex vivo identification of disease localizations in high-risk prostate cancer patients undergoing surgery. However, the accurate analysis of radiotracer uptake would require time-consuming manual volumetric segmentation of 3D images. The aim of this study was to test the feasibility of using machine learning to perform automatic nodal segmentation of intraoperative ⁶⁸Ga-PSMA-11 PET/CT specimen images. Six (n = 6) lymph-nodal specimens were imaged in the operating room after an e.v. injection of 2.1 MBq/kg of ⁶⁸Ga-PSMA-11. A machine learning-based approach for automatic lymph-nodal segmentation was developed using only open-source Python libraries (Scikit-learn, SciPy, Scikit-image). The implementation of a k-means clustering algorithm (n = 3 clusters) allowed to identify lymph-nodal structures by leveraging differences in tissue density. Refinement of the segmentation masks was performed using morphological operations and 2D/3D-features filtering. Compared to manual segmentation (ITK-SNAP v4.0.1), the automatic segmentation model showed promising results in terms of weighted average precision (97-99%), recall (68-81%), Dice coefficient (80-88%) and Jaccard index (67-79%). Finally, the ML-based segmentation masks allowed to automatically compute semi-quantitative PET metrics (i.e., SUVmax), thus holding promise for facilitating the semi-quantitative analysis of PET/CT images in the operating room.

SUBMITTER: Rovera G

PROVIDER: S-EPMC10529304 | biostudies-literature | 2023 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative <sup>68</sup>Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

Rovera Guido G Grimaldi Serena S Oderda Marco M Finessi Monica M Giannini Valentina V Passera Roberto R Gontero Paolo P Deandreis Désirée D

Diagnostics (Basel, Switzerland) 20230921 18

High-resolution intraoperative PET/CT specimen imaging, coupled with prostate-specific membrane antigen (PSMA) molecular targeting, holds great potential for the rapid ex vivo identification of disease localizations in high-risk prostate cancer patients undergoing surgery. However, the accurate analysis of radiotracer uptake would require time-consuming manual volumetric segmentation of 3D images. The aim of this study was to test the feasibility of using machine learning to perform automatic no ...[more]

PMID: 37761380

Similar Datasets

Project description:BackgroundRoutine prostate-specific membrane antigen (PSMA) positron emission tomography (PET) performed for primary staging or restaging of prostate cancer patients is usually done as a single static image acquisition 60 min after tracer administration. In this study, we employ dynamic whole-body (D-WB) PET imaging to compare the pharmacokinetics of [68Ga]Ga-PSMA-11 and [18F]PSMA-1007 in various tissues and lesions, and to assess whether Patlak parametric images are quantitative and improve lesion detection and image readability.MethodsTwenty male patients with prostate cancer were examined using a D-WB PSMA PET protocol. Ten patients were scanned with [68Ga]Ga-PSMA-11 and ten with [18F]PSMA-1007. Kinetic analyses were made using time-activity curves (TACs) extracted from organs (liver, spleen, bone, and muscle) and lesions. For each patient, three images were produced: SUV + Patlak parametric images (Ki and DV). All images were reviewed visually to compare lesion detection, image readability was quantified using target-to-background ratios (TBR), and Ki and DV values were compared.ResultsThe two PSMA tracers exhibited markedly different pharmacokinetics in organs: reversible for [68Ga]Ga-PSMA-11 and irreversible for [18F]PSMA-1007. For both tracers, lesions kinetics were best described by an irreversible model. All parametric images were of good visual quality using both radiotracers. In general, Ki images were characterized by reduced vascular signal and increased lesion TBR compared with SUV images. No additional malignant lesions were identified on the parametric images.ConclusionD-WB PET/CT is feasible for both PSMA tracers allowing for direct reconstruction of parametric Ki images. The use of multiparametric PSMA images increased TBR but did not lead to the detection of more lesions. For quantitative whole-body Ki imaging, [18F]PSMA-1007 should be preferred over [68Ga]Ga-PSMA-11 due to its irreversible kinetics in organs and lesions.

Project description:BackgroundProstate-specific membrane antigen (PSMA) PET/CT imaging is widely used for quantitative image analysis, especially in radioligand therapy (RLT) for metastatic castration-resistant prostate cancer (mCRPC). Unknown features influencing PSMA biodistribution can be explored by analyzing segmented organs at risk (OAR) and lesions. Manual segmentation is time-consuming and labor-intensive, so automated segmentation methods are desirable. Training deep-learning segmentation models is challenging due to the scarcity of high-quality annotated images. Addressing this, we developed shifted windows UNEt TRansformers (Swin UNETR) for fully automated segmentation. Within a self-supervised framework, the model's encoder was pre-trained on unlabeled data. The entire model was fine-tuned, including its decoder, using labeled data.MethodsIn this work, 752 whole-body [68Ga]Ga-PSMA-11 PET/CT images were collected from two centers. For self-supervised model pre-training, 652 unlabeled images were employed. The remaining 100 images were manually labeled for supervised training. In the supervised training phase, 5-fold cross-validation was used with 64 images for model training and 16 for validation, from one center. For testing, 20 hold-out images, evenly distributed between two centers, were used. Image segmentation and quantification metrics were evaluated on the test set compared to the ground-truth segmentation conducted by a nuclear medicine physician.ResultsThe model generates high-quality OARs and lesion segmentation in lesion-positive cases, including mCRPC. The results show that self-supervised pre-training significantly improved the average dice similarity coefficient (DSC) for all classes by about 3%. Compared to nnU-Net, a well-established model in medical image segmentation, our approach outperformed with a 5% higher DSC. This improvement was attributed to our model's combined use of self-supervised pre-training and supervised fine-tuning, specifically when applied to PET/CT input. Our best model had the lowest DSC for lesions at 0.68 and the highest for liver at 0.95.ConclusionsWe developed a state-of-the-art neural network using self-supervised pre-training on whole-body [68Ga]Ga-PSMA-11 PET/CT images, followed by fine-tuning on a limited set of annotated images. The model generates high-quality OARs and lesion segmentation for PSMA image analysis. The generalizable model holds potential for various clinical applications, including enhanced RLT and patient-specific internal dosimetry.

Project description:PurposeWe aimed to compare the diagnostic performance and biodistribution of two similar PET agents, [68Ga]Ga-P16-093 and [68Ga]Ga-PSMA-11, in the same group of primary prostate cancer (PCa) patients.MethodsFifty patients with untreated, histologically confirmed PCa by needle biopsy were enrolled. Each patient underwent [68Ga]Ga-P16-093 and [68Ga]Ga-PSMA-11 PET/CT within a week. In addition to visual analysis, the standardized uptake value (SUV) was measured for semiquantitative comparison and correlation analysis.Results[68Ga]Ga-P16-093 PET/CT detected more positive tumors than [68Ga]Ga-PSMA-11 PET/CT (202 vs. 190, P = 0.002), both for intraprostatic lesions (48 vs. 41, P = 0.016) and metastatic lesions (154 vs. 149, P = 0.125), especially for intraprostatic lesions in low- and intermediate-risk PCa patients (21/23 vs. 15/23, P = 0.031). Furthermore, [68Ga]Ga-P16-093 PET/CT exhibited a significantly higher SUVmax for most matched tumors (13.7 ± 10.2 vs. 11.4 ± 8.3, P < 0.001). For normal organs, [68Ga]Ga-P16-093 PET/CT showed significantly lower activity in the kidney (SUVmean: 20.1 ± 6.1 vs. 29.3 ± 9.1, P < 0.001) and urinary bladder (SUVmean: 6.5 ± 7.1 vs. 20.9 ± 17.4, P < 0.001), but displayed a higher uptake in the parotid gland (SUVmean: 8.7 ± 2.6 vs. 7.6 ± 2.1, P < 0.001), liver (SUVmean: 7.0 ± 1.9 vs. 3.7 ± 1.3, P < 0.001), and spleen (SUVmean: 8.2 ± 3.0 vs. 5.2 ± 2.2, P < 0.001) than [68Ga]Ga-PSMA-11 PET/CT.Conclusion[68Ga]Ga-P16-093 PET/CT demonstrated higher tumor uptake and better tumor detectability than [68Ga]Ga-PSMA-11 PET/CT, especially in low- and intermediate-risk PCa patients, which indicated that [68Ga]Ga-P16-093 may serve as an alternative agent for detection of PCa.Trial registration68Ga-P16-093 and 68Ga-PSMA-11 PET/CT Imaging in the Same Group of Primary Prostate Cancer Patients (NCT05324332, Registered 12 April 2022, retrospectively registered). URL OF REGISTRY: https://clinicaltrials.gov/ct2/show/NCT05324332 .

Project description:Baseline uptake on prostate-specific membrane antigen (PSMA)-targeted imaging is a prerequisite for radioligand therapy (RLT) with [177Lu]Lu-PSMA-617. This study aims to quantify lesion-based response to RLT in relation to pretreatment standard molecular imaging metrics derived from [68Ga]Ga-PSMA-11 PET/CT. Sixty-one patients with mCRPC underwent [68Ga]Ga-PSMA-11 PET/CT imaging before and after a median of 4 (IQR 2-6) RLT cycles. Maximum and mean standardized uptake values (SUVmax, SUVmean), as well as tumor-to-liver ratio (TLR), were assessed. A median of 12 (IQR 7-17) lesions was analyzed per patient, resulting in a total of 718 lesions. Lesions with ≥30% SUVmax decline or falling below the blood pool uptake were considered responsive; ≥30% SUVmax increase marked lesion progression. Additionally, 4-point visual scoring was performed according to E-PSMA consensus. In total, 550/718 (76.6%) lesions responded to RLT, including 389/507 (76.7%) bone metastases and 143/181 (79.0%) lymph node metastases. Baseline SUVmax, SUVmean, and TLR values were associated with lesion response by a moderate but significant correlation (rs = 0.33, p < 0.001, rs = 0.32, p < 0.001, and rs = 0.31, p < 0.001, respectively). For the classification of lesion progression based on baseline PSMA uptake, receiver operating characteristics (ROC) found SUVmax, SUVmean, and TLR to have comparable discriminatory value (AUC 0.85, 0.87, and 0.83). Of 42 tumor sites with baseline uptake below the liver (V-score < 2), 19/42 (45.2%) were responsive, 9/42 (21.4%) were stable, and 14/42 (33.3%) showed progression, leaving liver uptake a threshold with low prognostic value for the identification of RLT-refractory lesions (PPV 33%). This was observed accordingly for various liver uptake-based thresholds, including TLR < 1.5, <2.0 with a PPV at 24%, 20%, respectively. Standard uptake parameters quantified by routine baseline [68Ga]Ga-PSMA-11 PET/CT are moderately associated with post-treatment lesion response to [177Lu]Lu-PSMA-617. Commonly applied liver-based uptake thresholds have limited value in predicting refractory lesions at individual tumor sites.

Project description:Prostate-specific membrane antigen (PSMA) is a validated target for molecular diagnostics and targeted radionuclide therapy. Our purpose was to evaluate PSMA expression in hepatocellular carcinoma (HCC), cholangiocarcinoma (CCA), and hepatic adenoma (HCA); investigate the genetic pathways in HCC associated with PSMA expression; and evaluate HCC detection rate with 68 Ga-PSMA-11 positron emission tomography (PET). In phase 1, PSMA immunohistochemistry (IHC) on HCC (n = 148), CCA (n = 111), and HCA (n = 78) was scored. In a subset (n = 30), messenger RNA (mRNA) data from the Cancer Genome Atlas HCC RNA sequencing were correlated with PSMA expression. In phase 2, 68 Ga-PSMA-11 PET was prospectively performed in patients with treatment-naïve HCC on a digital PET scanner using cyclotron-produced 68 Ga. Uptake was graded qualitatively and semi-quantitatively using standard metrics. On IHC, PSMA expression was significantly higher in HCC compared with CCA and HCA (P < 0.0001); 91% of HCCs (n = 134) expressed PSMA, which principally localized to tumor-associated neovasculature. Higher tumor grade was associated with PSMA expression (P = 0.012) but there was no association with tumor size (P = 0.14), fibrosis (P = 0.35), cirrhosis (P = 0.74), hepatitis B virus (P = 0.31), or hepatitis C virus (P = 0.15). Overall survival tended to be longer in patients without versus with PSMA expression (median overall survival: 4.2 vs. 1.9 years; P = 0.273). FGF14 (fibroblast growth factor 14) mRNA expression correlated positively (rho = 0.70; P = 1.70 × 10-5 ) and MAD1L1 (Mitotic spindle assembly checkpoint protein MAD1) correlated negatively with PSMA expression (rho = -0.753; P = 1.58 × 10-6 ). Of the 190 patients who met the eligibility criteria, 31 patients with 39 HCC lesions completed PET; 64% (n = 25) lesions had pronounced 68 Ga-PSMA-11 standardized uptake value: SUVmax (median [range] 9.2 [4.9-28.4]), SUVmean 4.7 (2.4-12.7), and tumor-to-liver background ratio 2 (1.1-11). Conclusion: Ex vivo expression of PSMA in neovasculature of HCC translates to marked tumor avidity on 68 Ga-PSMA-11 PET, which suggests that PSMA has the potential as a theranostic target in patients with HCC.

Dataset Information

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative ⁶⁸Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

Publications

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative <sup>68</sup>Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets

Dataset Information

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative 68Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

Publications

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative <sup>68</sup>Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets

Machine Learning CT-Based Automatic Nodal Segmentation and PET Semi-Quantification of Intraoperative ⁶⁸Ga-PSMA-11 PET/CT Images in High-Risk Prostate Cancer: A Pilot Study.