RPS 2007 - Innovations in prostate imaging

Author Block: B. Oerther¹, H. Engel¹, R. Strecker², T. Benkert², E. Weiland², F. Bamberg¹, M. Benndorf¹, J. Weiß¹, C. Wilpert¹; ¹Freiburg/DE, ²Erlangen/DE
Purpose: The aim of this study was to establish and evaluate image quality of an ultra-fast MRI screening protocol for prostate cancer in biopsy-naïve men regarding PI-RADSv2.1 classification in comparison to the standard multiparametric protocol.
Methods or Background: This prospective mono-institutional study included consecutive patients with suspected prostate cancer without prior biopsy. A PI-RADSv2.1 conform multiparametric MRI protocol was acquired in a 3 T MRI scanner (scan time: 25min 45sec). Additionally, two deep-learning accelerated sequences were acquired (scan time: 3 min 28 sec). Two readers evaluated image quality and the presence of prostate cancer. In a first reading session only the screening protocol (DL accelerated axial T2w and ZOOMit DWI) was available. Subsequently, the full conventional mpMRI protocol was assessed and served as a reference standard. Diagnostic performance was analyzed with mpMRI serving as the gold standard. Inter- and intra-reader agreement was assessed using weighted kappa statistics.
Results or Findings: The final cohort consisted of 77 patients with 97 lesions. Diagnostic performance of the screening protocol was excellent with a sensitivity and specificity of 100%/97% and 98%/83% (cut-off ≥ PI-RADS 3) vs. 100%/100% and 89%/98% (cut-off ≥ PI-RADS 4) for reader 1 and reader 2, respectively. Mean image quality (Likert-scaling) was 3.96 (R1) and 4.35 (R2) for the standard protocol vs. 4.74 and 4.57 for screening protocol (p < 0.05). Inter-reader agreement was moderate (κ: o.55) for the screening protocol and substantial (κ: o.61) for the multiparametric protocol. Intra- reader agreement was excellent (κ: o.98) for R1 and substantial (κ: o.79) for R2.
Conclusion: A DL accelerated screening protocol for prostate cancer in biopsy-naïve men proved similar diagnostic performance and better imaging quality compared to the conventional mpMRI protocol, requiring less than 15% of scan time.
Limitations: Monocentric study, limited number of patients; no histopathological ground truth
Funding for this study: No funding was obtained for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was ethically approved the Local ethics committee of Freiburg.

Author Block: N. Sushentsev¹, L. A. Abrego Rangel², N. Sanmugalingam³, I. Caglič¹, V. Gnanapragasam¹, A. Warren¹, A. Zaikin², T. Barrett¹, O. Blyuss²; ¹Cambridge/UK, ²London/UK, ³Nowich/UK
Purpose: The aim of this study was to develop a dynamic predictive model for baseline detection and follow-up re-evaluation of the risk of prostate cancer (PCa) progression on active surveillance (AS).
Methods or Background: Four hundred and twenty two AS patients were included in this study, of whom 82 (19.4%) experienced either histological PCa progression or radiological stage progression (PRECISE 5) over a median follow-up of 4.5 years. The baseline model included initial serum prostate-specific antigen (PSA) and PSA density (PSAD), MRI-derived Likert score, tumour diameter, and tumour grade group. The follow-up model included baseline Likert score along with longitudinal PRECISE scores, PSAD measurements, and repeat biopsy results. Model training and testing were performed in the 50/50 data split using several neural networks, with three-year progression as the outcome.
Results or Findings: The best-performing baseline model was a generalised additive model (GAM) including baseline PSAD and Likert score. With an overall test AUC of 0.65, the model achieved a 21% specificity at 95% sensitivity in the test set, which may be used to avoid repeat biopsies in a substantial proportion of patients with minimal risk of missing disease progression. The follow-up model, comprised of a long short-term memory recurrent neural network, included baseline Likert score together with longitudinal PRECISE and PSAD measurements, with its test AUC of 0.75 being significantly higher compared to that of PRECISE alone (AUC=0.61, P<0.01).
Conclusion: The development of MRI-driven risk-adapted AS predictive models is a high research priority in the field; this study shows the promise of the proposed approach to objectively stratify patients at baseline and significantly improve the performance of current standard-of-care PRECISE assessment for detecting disease progression in the follow-up.
Limitations: Lack of external validation to be addressed in future studies.
Funding for this study: This study was funded by the Cancer Research UK ACED (A095792/EICEDAAP100009).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was ethically approved by the HRA and Health and Care Research Wales (HCRW), IRAS Project ID: 288185.

Author Block: B. Valentin, M. Boschheidgen, T. Ullrich, J. P. Radtke, P. Albers, G. Antoch, H-J. Wittsack, L. Schimmöller; Düsseldorf/DE
Purpose: This study aimed to improve the contrast between prostate cancer (PC) and healthy tissue by diffusion-weighted imaging (DWI) post-processing by using a square method.
Methods or Background: DWI post-processing was retrospectively applied on 40 patients with PC (median 68y and PSA 8.0 ng/ml) and multiparametric MRI (mpMRI) performed at 3 Tesla (Magnetom PRISMA® Siemens, Erlangen, Germany). In 20 patients a multi-shot readout segmentation (rs-EPI) plus zoomed single-shot imaging (z-EPI) sequence (Group 1) and in 20 patients a single-shot echo-planar imaging (ss-EPI) plus rs-EPI sequence (Group 2) was applied. All sequences (b1000 and b1800/2000) were squared and afterwards evaluated objectively using the SyngoVia Software (Siemens Healthineers, Erlangen) and subjectively by applying a 5-point scale from unacceptable, poor, moderate, good, to excellent.
Results or Findings: The squared processed DWI sequences showed significantly higher contrast-ratio (CR) for ss-EPI b1800 (p<0.001), rs-EPI b1000 (p<0.001), rs-EPI b1800 (p<0.001), z-EPI b1000 (p=0.002) and for z-EPI b2000 (p<0.001). Following post-processing of these sequences, a significant improvement in the subjective assessment of image quality was noticeable for ss-EPI b1000 (p=0.043), ss-EPI b1800 (p=0.030), rs-EPI b1000 (p<0.0001), rs-EPI b1800 (p<0.001) and z-EPI b1000 (p<0.001).
Conclusion: The application of the square post-processing for DWI results in a significant improvement in the CR between PC and healthy tissue, especially for high b values of ss-EPI or re-EPI. This method can be instrumental in enhancing the detection and differentiation of PC lesions.
Limitations: High b-values can cause overexposure of the lesion and highlighting of non-specific DWI restrictions. This could result in overdiagnosis as well as misinterpretation of the tumour lesion margins.
Funding for this study: This study did not receive any funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The local ethics committee has positively reviewed this retrospective study.

Author Block: B. Valentin, T. Thiel, T. Ullrich, M. Boschheidgen, J. P. Radtke, P. Albers, G. Antoch, H-J. Wittsack, L. Schimmöller; Düsseldorf/DE
Purpose: In this study, we applied the model-free nonnegative least squares (NNLS) method. The aim of this study was to assess the number of distinguishable diffusion components (spectral diffusion analysis) within the prostate and to differentiate between healthy and pathological prostate tissue.
Methods or Background: NNLS imaging was performed at 3 Tesla (Magnetom PRISMA® Siemens, Erlangen, Germany) in 10 patients with prostate cancer (PC) (PI-RADS 5 and subsequent biopsy). The 16 b values used were 0, 50, 100, 150, 200, 300, 400, 500, 600, 700, 800, 1000, 1200, 1400, 1600, 1800 s/mm 2. Relative signal fractions and mean diffusivities of the diffusion components in the peripheral zone, central zone and PI-RADS 5 lesion were obtained using the regularized NNLS fitting of the intravoxel incoherent motion data.
Results or Findings: Three different diffusion components (10–4, 10–3, and 10–2 mm2/s) were detected in prostate tissue. In comparison, the three peaks were significantly different between healthy and diseased tissue. The fraction of the slow component was significant higher in PC (maximum amplitude of 0.2) compared with the unaffected prostate tissue (maximum amplitude of 0.05).
Conclusion: This pilot study demonstrated the feasibility of spectral diffusion weighted imaging for the differentiation of PC. The three distinguishable components may be related to slow tissue diffusion caused by higher tissue density of PC lesions, intermediate fluid flow caused by glandular tissue, and fast blood flow in prostatic vessels. A larger cohort study with a ISUP range is needed to further evaluate this technique.
Limitations: This study exclusively focused on PI-RADS 5 lesions. Consequently, we were unable to differentiate between lesion sizes and did not account for variations in high-, intermediate-, and low-risk prostate cancer cases.
Funding for this study: This study was funded by DFG.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The local ethics committee has positively reviewed this prospective study.

Author Block: K. L. Lee, D. Kessler, T. Barrett; Cambridge/UK
Purpose: Large language models, such as ChatGPT and Bard, have sparked a wave of enthusiasm for their potential applications in clinical radiology, including formulating clinical interpretation of reports. This study aims to compare the classification abilities of ChatGPT, Bard, and two uroradiologists in assigning PI-RADS categories based on clinical text reports.
Methods or Background: Clinical prostate MRI text reports from 100 consecutive treatment-naïve patients undergoing mpMRI between 06.11.2022 to 28.12.2022 were analysed. Clinical history and concluding remarks were removed from the text reports. Two uroradiologists with 14 and 3 years of prostate MRI reporting experience, retrospectively independently classified PI-RADS 2.1 categories on the edited text reports. The same reports were inputted manually into online ChatGPT-3.5 and Bard platforms to generate PI-RADS classifications (without prior training). Original report classifications were considered definitive, and comparisons were made to compare the original reports, the two radiologists, ChatGPT, and Bard. Agreement rates and Κappa scores were analysed.
Results or Findings: In the original reports, 52/100 MRIs were classified as PI-RADS 2, 9/100 as PI-RADS 3, 19/100 as PI-RADS 4, and 20/100 as PI-RADS 5, respectively. Compared to the original classifications, the senior and junior radiologists concurred on 95% and 90% of the reports, respectively, while ChatGPT and Bard aligned both on 67 reports. Notably, Bard assigned a non-existent PI-RADS 6 classification to two patients (2%). The interreader agreement (Κ) between the original reports and the senior radiologist, the junior radiologist, ChatGPT, and BARD were 0.92, 0.85, 0.55, and 0.49, respectively.
Conclusion: Concordance on PI-RADS scoring was high among radiologists, however, ChatGPT and Bard demonstrate poor performance for the text-based classification task.
Limitations: The limitation of the study is that this a relatively small sample of 100 reports.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by Cambridge University NHS Foundation Trust (reference number: 288185).

Author Block: F. Tollens, F. G. Zöllner, A. Adlung, S. O. Schönberg, D. Nörenberg; Mannheim/DE
Purpose: The aim of this study was to quantify tissue sodium concentration (TSC) in healthy prostate tissue and prostate cancer regions based on internal references in order to evaluate TSC as a potential quantitative imaging biomarker.
Methods or Background: Thirty-six patients with clinically suspected prostate cancer were included into this prospective study and underwent clinical multiparametric magnetic resonance imaging (MRI) and additional sodium MRI of the prostate. Imaging was performed at 3T using a dual-tuned 1H/23Na body-coil to acquire a 3D radial density-adapted 23Na sequence. For the quantification of tissue sodium concentration, femoral blood vessels were chosen as an internal reference and reported sodium levels of ~81 mM were assumed. Peripheral zone (PZ), transition zone (TZ) and tumor regions of interest were defined and TSC was extracted for each segmentation.
Results or Findings: Mean TSC differences between right and left femoral blood vessels was 3.3 ± 2.2 mM. TSC was significantly higher in the PZ (40.7 ± 6.0 mM) than in the TZ (37.5 ± 5.7 mM). Nine suspicious lesions (PI-RADS 4 and 5) were detected in eight men that were confirmed with Gleason scores of ≥ 3+3 by biopsy. TSC in prostate cancer (32.2 ± 5.5 mM) was significantly lower than in contralateral healthy regions (36.1 ± 3.9 mM, p=0.018).
Conclusion: Femoral blood vessels as an internal reference for TSC quantification are less prone to inaccuracies caused by B1 inhomogeneities as opposed to external sodium probes, which enables a robust quantification. TSC was significantly decreased within prostate cancer compared to healthy prostate tissue. Reduced TSC could represent a quantitative imaging biomarker that could improve prostate cancer risk stratification.
Limitations: Patients’ blood sodium concentration assumed based on literature; only PI-RADS 4 and 5 lesions with histopathologic confirmation considered as confirmed prostate cancer lesions.
Funding for this study: This research project is part of the Research Campus M2OLIE and funded by the German Federal Ministry of Education and Research (BMBF) within the Framework “Forschungscampus: public-private partnership for Innovations” under the funding code 13GW0388A and 13GW0092D.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the local ethics committee.

Author Block: Y. Lefebvre¹, E. Venetis¹, T. Metens¹, T. Benkert², E. Weiland², M. A. Bali¹, N. Coquelet¹; ¹Brussels/BE, ²Erlangen/DE
Purpose: The goal of this study was to compares a conventional diffusion-weighted imaging (DWI-STD) to a faster DWI using deep-learning reconstruction strategies (DWI-DLR) for prostate cancer (PCa) imaging in terms of image quality (IQ) and diagnostic interpretation.
Methods or Background: From February 2023 to July 2023, we query a retrospective monocentric database of patients with prostatic lesions who underwent a 3T magnetic resonance exam. Each patient successively underwent the DWI-DLR (acquisition time: 3 min38 s; research application) and the DWI-STD (4 min58s). Image analysis was performed by an expert radiologist in prostate imaging. Using high acquired b-value DWI images (800 s/mm2) and apparent diffusion coefficient (ADC), each sequence was rated using a 5-point Likert scale for overall IQ, noise, sharpness, contrast, artifacts and distortion. Lesion characteristics were assessed based on calculated b-value DWI (1400 s/mm2) and ADC maps, and rated for diagnostic confidence and detectability. Rating comparisons between the DWI sequences were performed using receiving operating characteristic curves and associated areas under the curve (AUC). P-values testing the null hypothesis that AUC equal 0.5 were computed and p-values below 0.05 Bonferroni-corrected for multiple comparisons were deemed significant.
Results or Findings: Twenty patients were included (mean age: 66.9 years, range: 58–80 years). We found that overall IQ (b800 and ADC); noise, sharpness and distortion (b800) were better for DWI-STD. Noise, sharpness and distortion (ADC); contrast and artifacts (b800 and ADC) were better for DWI-DLR. For lesions, diagnostic confidence and lesion detectability were better for DWI-DLR compared to DWI-STD. Significance was reached for contrast (b800 and ADC), noise and sharpness (ADC), and lesion detectability (b1400).
Conclusion: Our preliminary results show that PCa DWI can be acquired more rapidly using deep-learning reconstruction strategies without loss of IQ and diagnostic interpretation.
Limitations: No limitations were identified.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: The study is retrospective.

Author Block: S. Schirò, L. Leo, M. Russomando, P. Canale, C. Zilioli, V. Casale, C. Roberti, C. Manna, N. Sverzellati; Parma/IT
Purpose: The purpose of this study was to assess the diagnostic performance of MRI radiomic model in the prediction of anterior zone prostate cancer.
Methods or Background: This study included a retrospective monocenter dataset of subjects with prostate 1.5 T mpMRI showing an anterior zone lesion (PIRADS 3-5) and fusion target biopsy within six months. The histopathology results were the standard of reference. Two radiologists ( four years of experience) independently, reviewed and manually segmented the lesion with an open-source software at T2-weighted and ADC maps. Overall, 851 radiomics features (FR) were extracted for both T2-weighted and ADC maps. 100 train:test (0.7:0.3) splits were created and recursive feature elimination with a 5-fold cross-validation was performed on train partitions using the Random Forest Classifier (RFC). Subsequently, RFC was trained by selecting iteratively an increasing number of features sorted by their occurrences to evaluate the minimum number of informative features. Finally, means and 95% confidence intervals of accuracy, sensitivity, specificity, precision, area under the receiver operating characteristic curve (ROC-AUC) were calculated on the test partitions.
Results or Findings: Overall, 89 males (mean age 68 years; ±SD 8) were included. The anterior zone lesions were scored as follows: PI-RADS 3 (n=28, 31%), PI-RADS 4 (n=40, 45%) and PI-RADS 5 (n=21, 24%). Of these, 47/89 (53%) showed anterior zone prostate cancer (Gleason Score ≥ 3+3). The best model on the test set exploited six first-order features of ADC maps and three first-order features of T2-weighted images reaching accuracy, sensitivity, specificity, precision and ROC-AUC of 0.74 [95% C.I. 0.73-0.75], 0.72 [0.70-0.75], 0.77 [0.74-0.80], 0.80 [0.78-0.82], 0.82 [0.80-0.83], respectively.
Conclusion: The proposed radiomic model reached satisfactory performance in predicting anterior zone prostate cancer and may be a useful supportive tool in the diagnostic pathway.
Limitations: Monocentric retrospective cohort and no external validation
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by CE AVEN : AVEN873/2018/OSS/UNIPR.

Author Block: C. Pizzi, C. Sattin, S. Alessi, Q. Nguyen, G. Musi, G. Petralia; Milan/IT
Purpose: The goal of this study was to evaluate the detection rates of prostate cancer (PCa) and clinically significant prostate cancer (csPCa) with in-bore MRI-guided prostate biopsy.
Methods or Background: We considered in-bore MRI-guided prostate biopsies performed at our institution from November 2014 to January 2023, using two different systems: one manual and one robotic. Differences in detection rates of PCa and csPCa were assessed between the two systems and between patients with lesions ≤10 mm and >10 mm.
Results or Findings: Of the 831 biopsies included, 493 were positive for PCa and 281 for csPCa (59.3% and 33.8%). The detection rate of PCa and csPCa did not differ between manual and robotic systems (55.3% versus 63.2% and 33.2% versus 34.4%, respectively), neither for lesions ≤10 mm and >10 mm (32.9% versus 25.5% and 18.1% versus 16%, respectively).
Conclusion: The observed detection rates of PCa and csPCa (59.3% and 33.8%) agree with the literature. In the absence of significant differences in detection rates of PCa and csPCa, MRI-guided bore biopsy can be performed effectively on lesions of any size using either commercially available system (manual or robotic).
Limitations: Not applicable.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This institutional review board approved study was a single-centre retrospective analysis, and the specific consent was waived for this audit of clinical procedures that had been performed with informed consent.

Author Block: B. Carreri, P. E. Gigliotti, F. R. Fraioli, M. Nezzo, F. G. Garaci, G. Manenti; Rome/IT
Purpose: The aim of this prospective interventional study is assessing the feasibility, safety and treatment success of US/ MRI fusion-guided transperineal laser ablation (TPLA) as salvage treatment for radiation therapy refractory focal PCa, evaluating clinical and functional outcomes and defining post-procedural imaging findings using 3T multi-parametric MRI.
Methods or Background: A cohort of five patients over 70 years old who had undergone RT as primary treatment, with a single, Gleason score ≤ 7 (3 + 4), local recurrence for PCa, underwent TPLA in outpatient setting (SoracteLite, ECHOLASER, Elesta). Post-ablation follow-up included regular PSA sampling and 3T mpMRI at one hour, three, six, 12, and 18 months and systematic and targeted ultrasound/MR fusion-guided biopsies at 18 months. International Prostate Symptom Score (IPSS) and the five-item version of the International Index of Erectile Function (IIEF-5) questionnaires were completed at baseline and at 18-month follow-up to investigate any procedure-related erectile dysfunction or urinary symptoms.
Results or Findings: All procedures were successfully completed with no significant complications (Clavien-Dindo Grade I). The procedure achieved optimal outcomes, with a statistically relevant reduction of PSA and ablation cavity volume trends at the end of follow-up (>70%). IIEF-5 and IPSS scores showed no significant difference between pre-procedural and 18 months values. Ultrasound/MRI fusion-guided biopsies at 18 months also confirmed the absence of recurrence.
Conclusion: Preliminary results demonstrated that TPLA can effectively and safely treat local recurrences of RT refractory PCa over a medium-term period, without side-effects and functional complications.
Limitations: The limitations of our study are that it’s single-centred and no preliminary sample size calculation was conducted. Instead, we only included a small cohort of patients who met our specific criteria, such as having received radiation therapy, having a low-risk score, no extraglandular extension.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved the ethics committee: UID RS 68/19.

Author Block: L. Lambert¹, M. Jurka¹, M. Wagnerová¹, O. Capoun¹, R. Jakubicek², P. Ourednicek², A. Burgetova¹; ¹Prague/CZ, ²Brno/CZ
Purpose: Deep-learning-based reconstruction (DLR) is being developed by major MRI vendors in order to improve image quality and accelerate image acquisition. We used DLR combined with compressed sensing (CS), a method of K-space subsampling, the aim was to accelerate acquisition of T2-weighted MRI of the prostate gland and maintain original image quality.
Methods or Background: In this prospective study, forty-seven patients underwent biparametric prostate MRI with two T2 acquisitions in the transverse plane- a standard acquisition (4:27 min) and short acquisition (2:18) accelerated by increasing the CS factor from 1.3 to 2.8. The images were reconstructed with and without DLR. The image quality was rated in six domains, contrast-to-noise ratio and image sharpness were measured.
Results or Findings: The image quality of short-DLR was rated better in all categories compared to the standard sequence (p<0.0001 to p=0.0044). DLR images had higher sharpness compared to non-DLR. Both short and short-DLR images had lower calculated CNR. Subjective evaluation correlated both with prostate volume and image sharpness (p<0.0001).
Conclusion: The combination of DLR and CS results in accelerated acquisition of T2 images of the prostate with maintained perceived image quality, higher sharpness, and lower contrast-to-noise ratio.
Limitations: Study limitations: single center, single vendor, optimization of one sequence, technical level study
Funding for this study: This study was supported by the Ministry of Health of the Czech Republic (MH CZ-DRO, General University Hospital in Prague, 00064165) and by the institutional funding of the Charles University in Prague (Cooperatio, Medical Diagnostics and Basic Medical Sciences).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Ethics Committee of the General University Hospital in Prague

Author Block: L. Luo, R. Chang, Y. Li, Y. Liao, Z. Wang, X. Duan; Xi’an/CN
Purpose: This study aims to investigate the utility of machine learning-based radiomics models derived from PSMA PET/CT images in differentiating between benign and malignant intraprostatic lesions detected by [18F]-PSMA-1007 PET/CT.
Methods or Background: We retrospectively analyzed consecutive patients with prostate cancer (PCa) who underwent [18F]-PSMA-1007 PET/CT imaging and biopsy. A total of 1316 radiomic features were extracted from volumes of interest on PET and CT images respectively. Feature selection was performed using the Max-Relevance and Min-Redundancy (mRMR) algorithm and the least absolute shrinkage and selection operator (LASSO). Two radiomics models (PET model and PET/CT model) were generated using logistic regression in the training set. In addition, two baseline models were developed using clinical data including fPSA and tPSA, as well as prostate cancer molecular imaging evaluation standards (PROMISE), namely Clinical model and PROMISE model. ROC and Delong test were employed for model evaluation and comparisons.
Results or Findings: A total of 75 patients (50 with PCa and 25 with benign prostate hyperplasia) were included, with 53 patients for training and 22 for testing. In the training set, the area under curve (AUC) of the PET and PET/CT models were 0.94 (95% CI: 0.88-1.00) and 0.97 (95% CI: 0.92-1.00), respectively. The best-performing model (PET/CT model) demonstrated an accuracy of 92.5%, sensitivity of 100%, and specificity of 77.8%. Although the AUC of the PET/CT model was not significantly better than that of the PET model, it was significantly outperformed the Clinical model and PROMISE model (P < 0.001 and P = 0.002, respectively.).
Conclusion: Our findings highlight the potential clinical relevance of [18F]-PSMA-1007 PET-based radiomics models in the non-invasively prediction of intraprostatic lesions in patients with PCa, and show better diagnostic performance compared to baseline models.
Limitations: Not applicable
Funding for this study: No funding was obtained for this study,
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approval was granted by the Ethics Committee of the First Affiliated Hospital of Xi’an Jiaotong University.