Search Results (255)

Large language models (LLMs) are increasingly explored for clinical use. However, the extent to which such models can reliably support physicians in reporting, staging, and the assessment of classification remains an active area of research. This study aimed to evaluate and compare multiple LLMs for automated PROMISE V2 classification for prostate cancer. A total of 126 unambiguous German-language PSMA PET/CT text reports were retrospectively analyzed, with reference standards established by expert consensus based on image interpretation and the original report text. Five LLMs (GPT-5.4, DeepSeek-V3.2, Claude Sonnet 4.6, Gemini 3 Flash and Grok 4) were assessed using two English-language prompting strategies of varying complexity. Agreement with the reference standard served as the primary endpoint. Performance varied in the short-prompt setting (36.5–79.4%) but improved consistently with the long prompt (74.6–86.5%), with Gemini 3 Flash achieving the highest agreement. Across PROMISE V2 subcategories, agreement rates were high (miT: 81.0–92.1%, miN: 92.9–96.0%, miM: 92.9–95.2%), despite inter-model differences. In conclusion, contemporary LLMs demonstrate promising performance in deriving PROMISE V2 scores from unambiguous original report texts, particularly when guided by detailed prompts. Full article

Background: Accurate survival prediction is essential for optimizing treatment planning in patients with castration-resistant prostate cancer (CRPC). However, traditional statistical models often underperform because of limited variable inclusion and an inability to account for complex, multidimensional data interactions. Methods: We retrospectively collected 46 clinical, laboratory, and pathological variables from 801 patients with CRPC, covering the disease course from initial diagnosis to CRPC progression. Multiple machine learning (ML) models, including random survival forests (RSF), XGBoost, LightGBM, and logistic regression, were developed to predict cancer-specific mortality (CSM), overall mortality (OM), and 2- and 3-year survival status. The dataset was divided into training and test cohorts (80:20), and 10-fold cross-validation was performed. Performance was assessed using the C-index for regression models and the area under the curve (AUC), accuracy, precision, recall, and F1-score for classification models. Model interpretability was evaluated using SHapley Additive exPlanations (SHAP). Results: Over a median follow-up of 24 months, 70.6% of patients experienced CSM. Although XGBoost with its own imputation method achieved the highest C-index in the validation set, RSF demonstrated more stable performance and achieved the highest C-index in the held-out test set for both CSM (0.772) and OM (0.771). For classification tasks, RSF demonstrated superior performance in predicting 2-year survival, whereas XGBoost achieved the highest F1-score for 3-year survival prediction. SHAP analysis identified time to first-line CRPC treatment, hemoglobin level, and alkaline phosphatase level as key predictors of survival outcomes. Conclusions: RSF demonstrated robust test-set performance for time-to-event prediction, whereas XGBoost showed complementary value for 3-year survival classification. These models provide accurate and interpretable prognostic tools that may support personalized treatment strategies. External validation and integration of emerging therapies are warranted to enhance broader clinical applicability. Full article

Background/Objectives: Ataxia telangiectasia-mutated kinase is encoded by the ATM gene. This gene is altered in many cancer types; however, it is most relevant for eligibility of poly-ADP-ribose polymerase inhibitor (PARPi) therapy in metastatic castrate-resistant prostate cancer in Canada. Non-benign ATM alterations are relatively uncommon, seen in less than 10% of prostate cancers; however, many of these are variants of uncertain significance with limited evidence of clinical significance or actionability. Methods: To aid in variant classification of next-generation sequencing results, ATM immunohistochemistry (IHC) is performed on patient tumour samples to serve as patient-specific functional evidence for ATM protein loss. Results: ATM IHC demonstrates strong concordance with known loss-of-function variants and copy number losses. Additionally, it can be used to clarify ATM protein loss in cases involving variants of uncertain significance (VUSs) or copy number deletions. Conclusions: Incorporation of ATM IHC into clinical testing for prostate cancer represents a relatively cost-effective orthogonal approach that can rescue poor-quality NGS results and clarify the functional impact of ATM VUSs on protein expression. Overall, this strategy may provide useful supportive evidence in prostate cancer biomarker testing and interpretation. Full article

Prostate cancer (PCa) is predominantly an acinar adenocarcinoma arising from the prostatic glandular epithelium, with tumor grade assessed using the International Society of Urological Pathology (ISUP) Grade Group classification, reflecting the degree of glandular differentiation and underlying molecular heterogeneity. PCa exhibits wide clinical behavior heterogeneity, ranging from indolent disease to aggressive forms with poor outcomes. Accurate prognostic assessment is, therefore, essential for guiding treatment selection and monitoring disease progression. This review examines recent advances in imaging and non-imaging biomarkers that contribute to improved risk stratification, treatment planning, and disease monitoring. Particular attention is given to multiparametric magnetic resonance imaging (mpMRI), whole-body magnetic resonance imaging (WB-MRI), positron emission tomography/computed tomography (PET/CT), positron emission tomography/magnetic resonance imaging (PET/MRI), computed tomography (CT), and transrectal ultrasound (TRUS), evaluated for their capacity not only to detect disease but also to predict recurrence, progression, and survival outcomes. In parallel, the prognostic role of non-imaging biomarkers is discussed, including the prostate-specific antigen (PSA), histopathological grading, biochemical and inflammatory biomarkers, as well as genomic classifiers and circulating tumor DNA (ctDNA). Emerging approaches such as radiomics, liquid-biopsy-derived molecular profiles, and artificial intelligence (AI)-based multimodal integration are highlighted for their potential to enhance individualized decision making. This review underscores the importance of combining imaging and molecular information to refine prognostic models and accelerate the translation of precision medicine in PCa. Full article

Background/Objectives: Risk-based breast cancer (BC) screening can provide tailored recommendations based on individual risk. We aimed to identify key predictors for BC risk stratification to inform implementation in screening programs. Methods: We estimated 10-year BC risks using BOADICEA v.6 (CanRisk) in 3753 women aged 40–70 with no cancer history from the PERSPECTIVE I&I cohort. The primary endpoint was risk reclassification, assessed as the proportion of women whose assigned 10-year risk category changed when using different risk factor combinations against a full multifactorial model including questionnaire-based risk factors (QRFs), polygenic score (PGS), mammographic density (MD), and pedigree-structured first- and second-degree family history (FH) of breast, ovarian, pancreatic and prostate cancer, including both affected and unaffected relatives. Relative risk thresholds were set as <1.5 (average), 1.5–2.7 (higher-than-average), and ≥2.7 (high), equivalent to the remaining lifetime risk categories of <15%, 15–25% and ≥25% for women aged 30 (the anchor) to age 80. We quantified individual-level reclassification flows by direction and magnitude. Results: Excluding PGS from risk calculations led to the highest overall reclassification. Using only the BC status in first- and second-degree relatives produced comparable risk classification to that of the full FH data that included breast, ovarian, prostate and pancreatic cancer (reclassification = 0.5%). However, collecting only affected relatives led to overestimation of risk. Excluding either PGS, MD or FH resulted in a greater proportion of reclassification among younger women. Adding the PGS to risk factors already collected in provincial screening programs reduced reclassification from 23% to ~13%. Conclusions: PGS, MD, QRFs and FH of BC in affected and unaffected first- and second-degree relatives are key for refining risk stratification. These findings provide real-world evidence on how incorporating different sets of risk factors, both those routinely collected in screening programs and those requiring additional data collection, affect individual-level risk classification amongst a population-based cohort, and how the impacts differ across age groups. While risk classification reflects model-based changes in estimated risk categories rather than direct evidence of mis-screening or clinical outcomes, comparison with the current eligibility criteria used to identify women at higher-than-average risk highlights the potential clinical value of a multifactorial risk assessment approach in ensuring more appropriate screening strategies. Full article

Background: The steady rise of prostate cancer in Saudi Arabia signals a critical public health shift that requires immediate investment in early detection and prevention to mitigate a future clinical crisis. Accurate diagnosis using multiparametric MRI and PI-RADS scoring remains challenging, as interpretations are highly experience-dependent and subspecialized radiologists are limited. Methods: To address this gap, this study introduces a novel Dual-Stream Attention Network designed to automate the classification of low-risk (PIRADS 2-3) versus high-risk (PIRADS 4-5) lesions from T2-weighted MRI. Leveraging a ResNet50 backbone, the architecture employs parallel streams for Local and Global Feature Processing, each enhanced by a Channel-Spatial Attention module to highlight diagnostically relevant regions. These features are integrated through a Cross-Stream Fusion mechanism and a gate-controlled Adaptive Feature Fusion module to optimize multi-scale information. The model was developed and validated on a regional dataset of 3850 images from Jazan Specialist Hospital and Prince Mohammed bin Naser Hospital. This research provides a standardized, high-precision diagnostic path tailored to the Saudi Arabian population, conducted under institutional review board approval (No. 25138). Results: The proposed dual-stream attention network achieved an accuracy of 97.8% on the validation set and 96.4% on the test set, demonstrating high performance and generalization capabilities in classifying prostate lesions from Saudi patient populations. Conclusions: The proposed dual-stream architecture with novel attention and fusion mechanisms demonstrates high effectiveness for prostate cancer classification from T2-weighted MRI in Saudi clinical settings. This represents the first deep learning model specifically trained and validated on Saudi Arabian prostate MRI data, with the potential to address the shortage of specialized expertise and improve diagnostic efficiency in the Kingdom. Full article

The automatic classification of ISUP (International Society of Urological Pathology) grade groups in prostate histopathological images remains challenging due to the high similarity between adjacent classes, class imbalance, and label noise. In this work, we propose a deep learning pipeline based on EfficientNet convolutional neural networks combined with a hybrid loss function that integrates ordinal regression and Focal Loss to better capture the ordered nature of ISUP grades. A noise-filtering strategy based on the entropy of predictions from multiple EfficientNet models was first applied to identify and remove high-uncertainty samples from the training set. The problem was then reformulated as an ordinal regression task to explicitly model the hierarchical relationship among grades. Experiments conducted on the PANDA dataset demonstrate that removing noisy samples improved performance from $\kappa =0.826$ to $\kappa =0.833$. Incorporating ordinal loss further increased performance to $\kappa =0.851$. The best configuration, combining ordinal regression and Focal Loss, achieved $\kappa =0.857$ and an accuracy of $0.669$, while reducing severe misclassifications and concentrating errors among adjacent classes. These results indicate that explicitly modeling ordinal structure and mitigating label noise are effective strategies for improving prostate cancer grading systems. Full article

Background/Objectives: Prostate cancer (PCa) incidence is rising, with radical prostatectomy (RP) as the main curative surgery for localised cases, which includes removing seminal vesicles (SV). SV invasion (SVI) predicts poor oncological outcomes, making accurate preoperative staging to identify SVI crucial for surgical planning. This ensures oncological safety by enabling wide excision when needed, while preserving tissue to maintain function. This review synthesises current evidence on pre-biopsy MRI findings and/or clinicopathological parameters to diagnose SVI in PCa. Methods: A literature search (2005–2025) using OVID for studies assessing pre-biopsy MRI findings, with a priori eligibility for clinicopathological or combined MRI–clinicopathological models (index tests), for detecting SVI (outcome) compared to RP histopathology (standard reference) in patients with primary localised PCa (patients). This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Risk of bias was evaluated with QUADAS-2, and pooled diagnostic accuracy metrics and study heterogeneity were analysed. Results: Five studies qualified, while three used binary mpMRI classification and were quantitatively analysed. No eligible studies assessed clinicopathological predictors or combined MRI–clinicopathological models; all included studies evaluated pre-biopsy MRI findings only, and none included high-dimensional radiomics. The pooled sensitivity was 0.66 (95% CI: 0.52–0.78), specificity 0.94 (0.89–0.97), positive predictive value (PPV) 0.76 (0.60–0.87), negative predictive value (NPV) 0.92 (0.85–0.94), and diagnostic odds ratio 30.13 (12.36–73.47), with moderate heterogeneity. All included studies were retrospective cohorts with considerable risk of bias. Conclusions: In the small number of heterogeneous, single-centre retrospective studies available, pre-biopsy MRI findings show high specificity and NPV for preoperative detection of SVI but only moderate sensitivity, which limits its reliability as a standalone tool. The pooled diagnostic accuracy estimates should be interpreted as exploratory. These findings should therefore be interpreted cautiously. Future studies must integrate MRI with clinicopathological data, addressing this key evidence gap before firm conclusions can be drawn or clinical practice changed. Full article

Metastatic castration-resistant prostate cancer (mCRPC) is a prevalent malignancy marked by molecular heterogeneity, which contributes to resistance to standard therapies and poor clinical prognosis. Advances in genomic and transcriptomic profiling have identified key drivers such as alterations in AR, TP53, PTEN, and RB1, which also enable cancer cells to circumvent therapies. Despite such advances, the underlying mechanisms involved in mCRPC drug resistance are complex, creating an urgent need for novel therapies to improve clinical outcomes. To address this clinical problem, strategies focused on targeting underlying molecular and metabolic supportive pathways using nano-delivery systems of diverse drugs could be promising in both CRPC and mCRPC therapy. This review provides an overview of the current understanding of the genomic and microenvironmental landscape of mCRPC and explores emerging classification frameworks aimed at improving patient outcomes. We highlight the potential of integrative multi-omics approaches to inform precision oncology and guide the development of more effective, personalized treatments for prostate cancer therapy. Full article

Background: Transperineal magnetic resonance (MRI)/ultrasound (US) fusion-guided prostate biopsy has emerged as a promising alternative to the transrectal approach by improving lesion targeting and reducing infectious complications. However, real-world data addressing factors that influence the detection of clinically significant prostate cancer (csPCa), including imaging characteristics and procedural experience, remain limited. Objective: To evaluate the diagnostic performance, safety profile, and independent predictors of csPCa detection in patients who underwent transperineal MR/US fusion-guided prostate biopsy, with particular emphasis on PIRADS category, prostate-specific antigen (PSA) level, and procedural learning curve. Methods: In this study, patient data were prospectively recorded in a routinely maintained institutional database, while the present analysis was conducted retrospectively. A total of 136 patients with clinical suspicion of prostate cancer—defined as elevated prostate-specific antigen (PSA), abnormal digital rectal examination, or PIRADS ≥3 on multiparametric MRI—underwent transperineal MR/US fusion-guided biopsy between January 2023 and October 2024. Results: Prostate cancer was detected in 45.5% of patients, whereas csPCa was identified in 32.3%. The PIRADS category emerged as the strongest independent predictor of csPCa detection, with PIRADS-5 lesions showing a significantly greater likelihood of csPCa than PIRADS-3 lesions (OR 6.70, p = 0.006). The PSA level was also independently associated with csPCa detection (OR 1.06 per ng/mL increase, p = 0.033). Although csPCa detection rates increased across learning curve groups, procedural experience was not an independent predictor after adjustment. The procedure demonstrated a favorable safety profile, with a low rate of infectious and noninfectious complications despite minimal use of antibiotic prophylaxis. The multivariable model showed moderate explanatory power and acceptable overall classification accuracy. Conclusions: Transperineal MR/US fusion-guided prostate biopsy provides reliable detection of clinically significant prostate cancer with a low complication rate and consistent performance across different stages of institutional experience. The PIRADS category and PSA level remain key determinants of csPCa detection, supporting the integration of MRI-based risk stratification into contemporary prostate cancer diagnostic methods. Full article

Background: Precise staging of prostate cancer is vital for treatment planning and prognosis. While [⁶⁸Ga]Ga-PSMA-11 PET-CT has demonstrated high diagnostic accuracy in detecting metastatic disease, the interpretation of indeterminate or potentially benign PSMA-avid bone lesions remains a clinical challenge in routine practice. Methods: We conducted a retrospective single-centre study involving 214 patients who underwent [⁶⁸Ga]Ga-PSMA-11 PET-CT between January 2021 and January 2024. Patients with prior known bone metastases or alternative PSMA radiotracers were excluded. Only those with follow-up imaging were included for diagnostic accuracy analysis. Follow-up modalities included PSMA PET-CT, CT, MRI, and bone scintigraphy. Final classification (metastatic or benign) was based on radiological and clinical assessment. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated using follow-up imaging as the reference standard. Lesions classified as indeterminate were analysed separately and excluded from diagnostic performance calculations. Results: Of the 214 included patients, 142 had follow-up imaging. Among 80 patients with bone lesions initially reported as metastatic, 74 (92.5%) were confirmed. Among 28 patients initially reported as having benign bone lesions, 26 (92.9%) remained benign on follow-up. Thirty-four patients with indeterminate lesions were reviewed; four were ultimately metastatic. Excluding indeterminate cases, sensitivity, specificity, PPV, and NPV were 97.4%, 86.7%, 94.9%, and 92.9%, respectively. Diagnostic discordance was primarily associated with benign uptake in the ribs, iliac bones, pubic rami and degenerative changes. Conclusions: [⁶⁸Ga]Ga-PSMA-11 PET-CT shows excellent sensitivity and positive predictive value for detecting metastatic bone disease in prostate cancer. However, benign lesions may also exhibit uptake, emphasising the importance of integrating imaging results with PSA levels, Gleason scores, and TNM staging. Prospective studies are needed to validate these findings and assess their impact on long-term outcomes. Full article

Background: Subtype classification for breast cancer (BC) patients is important for risk-stratification. Unfortunately, this parameter is not always able to discriminate between high- and low-risk diseases. Glutamate Carboxypeptidase-II (GCPII), also known as prostate-specific membrane antigen (PSMA), could be an important biomarker of aggressiveness, given that it has been reported to be expressed in BC tumor cells and even more in endothelial cells of tumor vessels. Methods: We analyzed 22 Luminal A, 47 Luminal B, 9 HER2-positive (HER2+), and 23 triple-negative (TN) BC to assess whether PSMA, Ki67 expression, and tumor-infiltrating lymphocytes (TILs) were different in BC subtypes. Results: Median PSMA and Ki67 values were significantly higher in TNBC than in Luminal A and B tumors. We saw a correlation between PSMA and Ki67 expression, especially in HER2+ tumors (p = 0.035), while an inverse correlation between PSMA and TILs was observed in Luminal A (p = 0.028). Conclusions: Our results suggest that PSMA could be used as a biomarker in BC, given that it is highly expressed in more aggressive tumors. These findings open the way to a clinical investigation for the possible use of PSMA as a theranostic biomarker in BC patients with PSMA positive PET scan. Full article

Bone scintigraphy is a widely available and cost-effective modality for detecting skeletal metastases in prostate cancer, yet visual interpretation can be challenging due to heterogeneous uptake patterns, benign mimickers, and a high reporting workload, motivating robust computer-aided decision support. In this study, we present an experimental evaluation of fourteen convolutional neural network (CNN) architectures for binary metastasis classification in planar bone scintigraphy using a unified protocol. Fourteen models, CNN (baseline), AlexNet, VGG16, VGG19, ResNet18, ResNet34, ResNet50, ResNet50-attention, DenseNet121, DenseNet169, DenseNet121-attention, WideResNet50_2, EfficientNet-B0, and ConvNeXt-Tiny, were trained and tested on 600 scan images (300 normal, 300 metastatic) from the Jordanian Royal Medical Services under identical preprocessing and augmentation with stratified five-fold cross-validation. We report mean ± SD for AUC-ROC, accuracy, precision, sensitivity (recall), F1-score, specificity, and Cohen’s κ, alongside calibration via the Brier score and deployment indicators (parameters, FLOPs, model size, and inference time). DenseNet121 achieved the best overall balance of diagnostic performance and reliability, reaching AUC-ROC 96.0 ± 1.2, accuracy 89.2 ± 2.2, sensitivity 83.7 ± 3.4, specificity 94.7 ± 2.2, F1-score 88.5 ± 2.5, κ = 0.783 ± 0.045, and the strongest calibration (Brier 0.080 ± 0.013), with stable fold-to-fold behaviour. DenseNet121-attention produced the highest AUC-ROC (96.3 ± 1.1) but exhibited greater variability in specificity, indicating less consistent false-alarm control. Complexity analysis supported DenseNet121 as deployable (~7.0 M parameters, ~26.9 MB, ~92 ms/image), whereas heavier models yielded only limited additional clinical value. These results support DenseNet121 as a reliable backbone for automated metastasis detection in planar scintigraphy, with future work focusing on external validation, threshold optimisation, interpretability, and model compression for clinical adoption. Full article

Background: Prostate-specific membrane antigen (PSMA) is markedly overexpressed in prostate cancer (PCa), and there is growing evidence to support its usefulness in initial diagnostic assessments. This study compares the diagnostic performance of PSMA positron emission tomography/computed tomography (PET/CT) and magnetic resonance imaging (mpMRI) in evaluating seminal vesicle invasion (SVI), extraprostatic extension (EPE), and pelvic lymph node involvement before radical prostatectomy. Methods: A retrospective, single-institution analysis was performed. From a cohort of 325 patients who underwent radical prostatectomy between June 2022 to November 2024, 85 had undergone preoperative PSMA PET/CT for intermediate- and high-risk disease at biopsy, forming our study group. Two blinded specialists, one in radiology and one in nuclear medicine, independently interpreted the scans, using histopathological results as the reference standard. The primary outcome was diagnostic accuracy for T- and N-stage classification, while the secondary outcomes included the correct identification of the index lesion and comparative performance for each modality. Results: The study cohort comprised patients with intermediate-to-high-risk prostate cancer (median age: 66 years; median PSA level: 11.6 ng/mL; median PSA density: 0.3 ng/mL/cm³). Forty-eight patients presented with an ISUP grade of 3 or higher on biopsy. PSMA PET/CT was more sensitive than MRI for detecting EPE (72.2% vs. 46.9%) and nodal metastases (91.7% vs. 8.3%). Furthermore, PSMA PET/CT demonstrated significantly higher concordance with histopathological findings in index tumor localization (76.5% vs. 67.9%, p < 0.001). An exploratory analysis revealed a potential age-dependent pattern, but this requires confirmation in larger studies. Conclusions: In this select cohort, PSMA PET/CT demonstrated greater accuracy than MRI for locoregional staging in patients with intermediate-to-high-risk prostate cancer (PCa). However, the generalizability of these findings is limited by the retrospective design and potential selection bias. These results suggest that PSMA PET/CT may have a valuable role in the initial staging workflow, but this needs to be confirmed in larger, prospective studies. An exploratory analysis suggested a potential age-dependent pattern, but this requires confirmation in larger studies. Full article

Extraprostatic extension (EPE) is an important prognostic factor in prostate cancer and influences nerve-sparing decisions during radical prostatectomy. Multiparametric MRI (mpMRI) is the standard for local staging, but its sensitivity for EPE remains limited, and its interpretation is subject to inter-reader variability. In this narrative review, we aim to create an overview of contemporary strategies for the preoperative detection of EPE. We searched PubMed, Embase, Web of Science, and Google Scholar, focusing on studies published between 2015 and 2025 including articles evaluating clinical parameters, mpMRI features, nomograms, radiomics, machine learning, and deep learning models for EPE prediction. The analyzed literature was compared with respect to diagnostic performance, validation strategy, and clinical applicability of individual methods. Clinical parameters and traditional nomograms provide moderate accuracy for EPE detection. mpMRI improves staging, with tumor–capsule contact length as the most important single imaging marker. Radiomics-based and machine-learning models matched and occasionally outperform conventional approaches, achieving AUC values ranging from 0.75 to 0.85. Deep-learning models demonstrated similar performance by directly analyzing imaging data, although most lacked external validation and were sensitive to dataset heterogeneity. Several radiomics and deep learning models demonstrated performance comparable to, and in selected studies exceeding, expert radiologist assessment. Binary EPE classification has limited clinical value, while side-specific and graded EPE assessment offers a more clinically relevant approach. Translation of these tools into routine practice will require multimodal, side-specific, and externally validated models supported by automated segmentation and explainable artificial intelligence frameworks. Full article