Introduction
Artificial intelligence (AI) refers to the ability of machines to perform tasks that typically require human cognition.1
The roots of AI can be traced back to the mid-20th century, notably through the pioneering work of Alan Turing.2 Since then, progress in computational power, data storage, and algorithmic development has enabled the practical implementation of AI.3 This so-called “fourth industrial revolution” is not only transforming industries but also challenging clinicians to rethink workflows and decision-making models in healthcare. Its increasing integration into medicine is reshaping diagnostics, treatment planning and surgical practice.4
Among medical specialties, oncologic urology is experiencing a remarkable technological transformation with AI playing a central role in this progress and contributing to improvements in patient care.5,6 AI focuses on creating algorithms that enable machines to reason and perform cognitive functions, such as problem-solving, object and language recognition and decision-making.7
Artificial intelligence, machine learning (ML), and deep learning (DL) can be conceptualized as nested matryoshka dolls.8 Traditional ML depends on manually designed features, like color or texture, while DL specifically focuses on large-scale neural networks, designed to process images, sounds, videos, and other complex data types;8 this capability allows DL to uncover complex patterns and details that are challenging for humans to detect.9
Artificial Neural Networks (ANNs) including Convolutional Neural Networks (CNNs) are key subfields of ML with CNNs representing a more advanced, domain-specific variation optimized for high-dimensional data such as images and videos.10
Despite continuous progress, urologic oncology still faces several critical unmet needs. These include reducing overdiagnosis and overtreatment, improving diagnostic and surgical precision,11 minimizing the learning curve for complex procedures,12 and supporting tailored clinical and therapeutic pathways. Equally important is enhancing patient education,13 reducing bureaucratic burdens on clinicians, and optimizing outpatient workflow and resource allocation.14,15
Artificial intelligence (AI) is uniquely positioned to address many of these challenges by analyzing complex datasets, identifying clinically relevant patterns, and assisting in both decision-making and communication.7
This narrative review addresses these challenges by outlining current AI applications in urologic oncology, critically assessing their limitations, and identifying concrete opportunities for clinical integration.
This article is intended as a narrative review rather than a systematic analysis. Accordingly, references were selected based on their clinical relevance, novelty, and impact in the field of urologic oncology, rather than through a structured PRISMA-based search strategy.
General Framework for AI in Urologic Oncology
Urologic cancers constitute a significant portion of the global cancer burden. According to GLOBOCAN 2020, prostate cancer ranks as the fourth most commonly diagnosed cancer worldwide, with approximately 1.4 million new cases, while bladder, with among 614 thousands, and kidney cancers, with among 435 thousands, are also among the top ten in incidence.16 In Italy, genitourinary cancers account for about 20% of all malignancies, with 2022 estimates indicating 40,500 new cases of prostate cancer, 29,200 of bladder cancer, 12,600 of kidney cancer, 2,300 of testicular cancer, and 500 of penile cancer.17
Urology is a medical specialty that has evolved rapidly throughout history and continues to advance by incorporating innovative technologies to improve patient care and outcomes.18
As a result of these advancements, AI is increasingly being used to diagnose and manage urological diseases. One of the key advantages of AI lies in its ability to efficiently and accurately analyze vast amounts of clinical and imaging data.19 Among genitourinary malignancies, artificial intelligence has been most extensively applied and studied in prostate, bladder, and renal cancers, where substantial clinical and imaging datasets are available. In contrast, the application of AI in testicular and penile cancers remains limited, reflecting the relative scarcity of data and lower incidence of these tumor types in the general population.
By identifying intricate patterns and subtle details in patient information, machine learning (ML) algorithms assist physicians in early detection, risk evaluation, and personalized treatment planning.19 Despite growing interest in AI applications for urologic cancers, current studies exhibit several important methodological limitations that hinder broader clinical translation. Many models are validated only on internal datasets, without external validation or comparison to traditional statistical approaches. In addition, substantial variability in data quality, training protocols and algorithm types further limits the reproducibility and generalizability of the results.
The reviewed studies demonstrate promising advances in the application of machine learning (ML) for the diagnosis, classification, and outcome prediction of urologic cancers. Many report high performance metrics, often exceeding those of traditional statistical models, particularly in image-based tasks. However, several limitations persist. A substantial proportion of studies lack external validation, rely on retrospective single-institution datasets, or use heterogeneous methodologies that complicate cross-comparison. Furthermore, few studies assess clinical implementation or cost-effectiveness.
Future research should prioritize multicenter collaborations, inclusion of non-Western populations, external validation of models, and direct comparisons between AI and conventional clinical tools. Addressing these gaps will be essential for translating ML applications into routine clinical practice.
Diagnosis and Medical Treatment of Main Urological Malignancies
Prostate Cancer
Prostate cancer (PCa) is the most common malignancy among men, accounting for 288,300 new cases (29% of all male cancers) and 34,700 deaths (11% of all cancer-related deaths) in the United States in 2023.20
Table 1 and Table 2 provide a comprehensive overview of the clinical pipeline of artificial intelligence applications in prostate cancer. Table 1 summarizes the main studies, including AI models, study designs, and performance metrics, while Table 2 illustrates the step-by-step integration of AI tools across the diagnostic and therapeutic pathway.
|
Table 1 Summary of Principal Articles Regarding AI Applied in Prostate Cancer
|
 |
Table 2 Pipeline of the Clinical Application of AI Tools in the Flow Chart of Diagnosis, Treatment and Therapy for Prostate Cancer
|
The diagnosis of PCa generally begins with a biopsy, with or without performing a prostate multiparametric MRI (mpMRI),30 prompted by initial suspicion due to elevated prostate-specific antigen (PSA) levels.
Arun Seetharaman et al developed the Stanford Prostate Cancer Network (SPCNet) which is a convolutional neural network (CNN) model designed to differentiate aggressive cancer, indolent cancer, and normal tissue on MRI.21 The model’s evaluation was conducted on both pixel and lesion levels in a cohort of 322 patients, including six with normal MRI scans and no cancer, 23 who had undergone radical prostatectomy, and 293 who underwent biopsy.21 It was trained using data from 78 patients who underwent radical prostatectomy and 24 individuals without prostate cancer.21 The model identified clinically significant lesions with an area under the receiver operating characteristic curve of 0.75 in radical prostatectomy patients and 0.80 in biopsy patients. Additionally, it detected up to 18% of lesions missed by radiologists, demonstrating sensitivity and specificity comparable to those of radiologists in identifying clinically significant cancer.21 This model holds promise for aiding physicians in precisely targeting the aggressive components of prostate cancer during biopsy or focal treatments.21
To confirm this thesis, Anindo Saha et al evaluated an AI system for detecting clinically significant prostate cancer on MRI, comparing its performance to radiologists using PI-RADS 2.1 and standard multidisciplinary care. The AI system was trained and externally validated using a dataset of over 10,000 MRI scans.22 In a test cohort of 400 cases, the AI outperformed radiologists with a higher area under the receiver operating characteristic (AUROCDAS (0.91 vs 0.86) and detected 6.8% more significant cancers at the same specificity.22 However, in a larger cohort of 1000 cases, it did not confirm non-inferiority to standard multidisciplinary readings.22 These results suggest AI could enhance prostate cancer diagnostics, but further clinical validation is required.22
Also Rafiqul Islam et al in his study explored machine learning-based approaches to detect prostate cancer using MRI images.23 It evaluated deep learning models (VGG16, VGG19, ResNet50, and ResNet50V2) for feature extraction and applied random forest classification to distinguish healthy from cancerous prostate tissues.23 ResNet50 achieved the highest accuracy (99.64%).23 However, this result was obtained using internal validation only, without testing on independent datasets, which limits its immediate applicability in clinical settings.23
Treatment strategies for localized PCa are determined based on the pathological analysis of biopsy samples. Pathologists diagnose prostate cancer and assign a Gleason grade group (GG) by examining hematoxylin and eosin (H&E)-stained tissue samples.31 AI has shown potential in diagnosing and grading prostate cancer, but its application has been limited by isolated studies with poor generalization across multinational cohorts, posing a major barrier to clinical integration. Bulten et al aimed to overcome these limitations by fostering collaborative development of AI algorithms for prostate biopsy analysis, ensuring reproducible and validated assessment across international datasets.32,33
To achieve this, the PANDA challenge was launched as the largest pathology competition to date.24 Algorithms trained on 10,616 prostate biopsy samples from multiple centers achieved expert-level accuracy, with concordance scores of 0.862 and 0.868 in two independent cross-continental validation sets, demonstrating AI models performance comparable to that of expert uropathologists.24
According to a work made by G. Bianchi et al AI integration, particularly using Graph Neural Networks (GNNs) and the Distilling Across Scales for MIL (DAS-MIL) algorithm, enhances tumor detection without pixel-level annotations.25 In his study of 506 biopsies, the model achieved an AUC of 0.93 and an accuracy of 0.86.25
AI can address a wide range of tasks beyond diagnosis and grading including measuring cancer length and volume,34 quantifying the percentage of Gleason pattern,35 identifying and assessing perineural invasion,36 quantifying immunohistochemistry (IHC) staining,37 and detecting and measuring cribriform patterns.38 The introduction of high-throughput automated pathology slide scanners is paving the way for the full digitization of pathology laboratories, expected to become a reality in the near future.39
AI can also enhance pathology laboratory workflows by identifying ambiguous cases that require additional IHC staining and automatically initiating these requests before pathologist review, thereby reducing turnaround times.14 Additionally, AI can perform fully automated quality control of Whole Slide Imaging (WSI) scans, detecting those that may need re-scanning or re-staining to improve quality.40 In a cutting-edge study, Rana et al explored the application of computationally generated H&E staining to unstained prostate biopsy images, showing that the algorithm-generated images could accurately replicate prostate tumor characteristics and be used for pathological diagnosis, thus facilitating the early detection of abnormalities in unstained tissue biopsies.26 Their work was a cross-sectional study in which conditional Generative Adversarial Networks (cGANs) were trained to convert unstained biopsy images into computationally H&E-stained images.26 The Structural Similarity Index Measure (SSIM) ranges from −1 to 1, where:
1 indicates that the two images are identical, 0 signifies no structural similarity and negative values indicate significant structural differences. The results obtained by Rana et al showed high similarity between computationally and physically H&E-stained images with a SSIM: 0.902.26 While the reverse-staining cGAN restored images to their original unstained form with comparable accuracy (SSIM: 0.900).26
Although AI has made significant advancements, it remains susceptible to potential flaws or biases that can result in missed or incorrect diagnoses; However, even AI-generated misdiagnoses can be beneficial.26 In the study of Tolkach et al, pathologists found false-positive alerts from AI useful, as these flagged areas that warranted closer examination and additional immunostaining for more accurate evaluation.27
There is a risk that pathologists might overly rely on AI predictions without critically evaluating them, raising concerns about potential overdependence and the compromise of diagnostic skills, which highlights the need for continuous monitoring of pathology labs using AI tools—similar to post-marketing surveillance for new drugs—to evaluate how pathologists interact with AI-generated results and to assess the broader impact of AI on clinical practice.39
Moreover, AI has shown significant potential in the field of medicine, particularly in the interpretation of medical images.
In the context of prostate cancer pathology, the US Food and Drug Administration (FDA) has authorized the use of Paige Prostate, a model designed to differentiate benign prostate biopsy samples from those indicative of suspected PCa.28 This model leveraged multiple instance learning with a recurrent neural network (MIL-RNN) to train on 12,132 whole slide images (WSIs), using pathology report diagnoses as labels; it achieved an impressive area under the curve (AUC) of 0.99 on the test set and 0.93 on an independent external validation set of over 12,000 slides.28
Treatment options for localized prostate cancer include active surveillance, ablative radiotherapy, and radical prostatectomy.41 Urinary incontinence and erectile dysfunction are among the most common adverse effects following robot-assisted radical prostatectomy (RARP).42,43 Hung et al developed a deep learning model (DeepSurv) to predict urinary continence recovery after RARP using automated performance metrics (APMs) collected from 100 RARPs and clinicopathological data, derived from computer-based data recording devices.29 Features were ranked and a heavier score was addressed to each one based on their importance in predicting continence. The top five ranked features were then used to classify surgeons into two groups: the four highest-scoring surgeons were categorized as “Group 1/APMs” (more efficient APMs), while the others were placed in “Group 2/APMs” (less efficient APMs).29 The study showed APMs as stronger predictors than clinicopathological features. In this study urinary continence was achieved in 79 patients (79%) after a median of 126 days.29 Particularly higher urinary continence rates at three and six months post-surgery were detected in patients classified as “Group 1/APM” (47.5% vs 36.7%, p=0.034, and 68.3% vs 59.2%, p=0.047, respectively).29 So this model is found to be a reliable predictor factor compared to clinicopathological features alone.29
Urothelial Cancer
Bladder cancer (BC) ranks as the seventh most frequently diagnosed cancer among men globally and the tenth most common when considering both sexes.44,45
Table 3 provides a summary of the principal studies evaluating the application of artificial intelligence in urothelial cancer, including study design, AI techniques, performance metrics, and main limitations.
 |
Table 3 Summary of Principal Articles Regarding AI Applied in Urothelial Cancer
|
Additionally, although less common, the occurrence of aggressive upper tract urothelial carcinoma is on the rise, leading to a growing number of cases involving locally advanced and high-grade tumors.53
AI plays a role in medical imaging, particularly computed tomography (CT) scans and magnetic resonance imaging (MRI), both of which are essential for UC diagnosis and prognosis.54 Several studies have demonstrated AI’s potentiality in this field. Xu et al developed machine learning algorithms based on radiomic features derived from multiparametric MRI (mpMRI) to differentiate bladder tumors from normal bladder wall.46 Similarly, Garapati et al employed morphological and textural features from CT urography to determine the stage of bladder cancer.47
AI could be a helpful tool in detecting bladder cancer and staging tumors. In fact, Ikeda et al used AI-assisted cystoscopy, which has demonstrated high accuracy in diagnosing bladder cancer by identifying subtle changes in the bladder wall.48 They developed a CNN-based system trained on 2102 cystoscopic images (1671 normal and 431 tumor lesions), achieving a sensitivity of 89.7% and a specificity of 94% in distinguishing normal tissue from tumor lesions.48
Moreover, AI-based analysis of cystoscopic images has demonstrated improved tumor clearance during transurethral resection of the bladder.55 The algorithm designed by Ibrahim Fahoum et al has the potential to significantly assist in accurately identifying muscularis propria invasion by replicating the analytical approach of a pathologist.49 This AI-driven approach could enhance the accuracy of bladder cancer diagnosis and treatment.56
The application of AI is not limited to just image analysis, but it could be also applied in genomics. AI-based studies aim to identify key genetic variants linked to UC, classify molecular subtypes of the disease, and explore pharmacogenomic approaches to determine which subtypes might respond to specific chemotherapy regimens.57,58 Most genomic studies utilize gene expression profiles or DNA/RNA sequencing data in AI models to predict prognosis, such as risks of disease progression, recurrence, or survival.57 However, the application of AI to predict treatment responses based on genetic data remains limited and further studies are warranted.
A systematic review and meta-analysis conducted by Chunlei He et al evaluated 21 studies that applied image-based AI models—including CT, MRI, radiomics, and deep learning—to predict muscle-invasive bladder cancer (MIBC), analyzing data from a total of 4,256 patients.50 Seventeen studies were included in the pooled analysis, which showed promising diagnostic accuracy (AUC up to 0.92 for MRI and 0.91 for deep learning models). Despite these encouraging results, the study highlighted substantial limitations in methodological and reporting quality, as assessed by CLAIM, RQS, and PROBAST, with all models carrying a high risk of bias.50
A systematic review and meta-analysis conducted by Caio Vinícius Suartz et al evaluated the performance of machine learning models in predicting response to neoadjuvant cisplatin-based chemotherapy (NAC) in patients with MIBC.51 From 12 studies (1,523 patients), four were included in the meta-analysis, reporting a sensitivity of 0.62 and specificity of 0.82. ML models using CT, genomic, and pathological data showed promising predictive ability, although further standardization is needed to improve reproducibility and clinical applicability.51
A multicenter study developed and validated AI-based histologic assays to predict recurrence, progression, BCG-unresponsive disease, and need for cystectomy in patients with high-risk non–muscle-invasive bladder cancer (NMIBC) treated with intravesical BCG.52 Using digitized pathology slides and clinical data from 944 patients across 12 centers, the AI models accurately stratified patients into high- and low-risk groups, providing prognostic information beyond traditional clinicopathologic factors.52 High-risk classifications were strongly associated with worse outcomes, including shorter recurrence-free and progression-free survival and higher likelihood of cystectomy.52
Renal Cancer
Renal cell carcinoma (RCC) is the sixth most frequently diagnosed cancer in men and the eighth in women worldwide, accounting for 5% of all cancer cases in men and 3% in women.59
Table 4 summarizes the key studies investigating the use of artificial intelligence in renal cancer, highlighting the type of AI models applied, study design, performance outcomes, and current limitations.
 |
Table 4 Summary of Principal Articles Regarding AI Applied in Renal Cancer
|
Many models have focused on differentiating benign from malignant kidney tumors, particularly distinguishing fat-poor angiomyolipomas from RCCs, achieving promising results.64 Fenstermaker et al developed a CNN model trained on 3,000 normal and 12,168 RCC tissue samples from 42 patients (digital H&E-stained images from The Cancer Genome Atlas) to identify RCC in histopathologic specimens, differentiate subtypes (ccRCC, chrRCC, pRCC), and predict Fuhrman grades The model achieved 99.1% accuracy for distinguishing normal tissue from RCC, 97.5% for RCC subtype classification, and 98.4% for Fuhrman grade prediction, demonstrating the potential of AI in RCC diagnosis, subtyping, and grading.60 Although the performance metrics were impressive, the model was developed using data from a limited number of patients and was not externally validated, which may reduce the robustness and generalizability of the findings.60
A noteworthy article in the field is the systematic review by Zine-Eddine Khene et al, which, by analyzing 20 studies, demonstrates the promising potential of machine learning and deep learning models in histological image analysis.65 However, they also emphasize the challenges of clinical implementation, including methodological heterogeneity, lack of external validation, and the limited interpretability of AI algorithms.65
In addition to this, AI could be applied in RCC histopathology, such as Alfredo Distante et al examine in their systematic review enhancing the use of AI in diagnostic precision, tumor grading, and prognosis prediction while reducing interobserver variability and pathologists’ workload.66 So AI shows promise in automating histological assessments and integrating molecular data for improved patient stratification.66 However, challenges remain regarding standardization, generalizability, and clinical implementation.66
Other AI-driven approaches explored, evaluate indirect markers of tumor aggressiveness, such as the SSIGN score (stage, size, grade, and necrosis), which predicts the progression of ccRCC following radical nephrectomy. For instance, Choi et al developed an AI algorithm to preoperatively predict low versus high SSIGN scores in ccRCC patients undergoing MRI, demonstrating a high AUC of 0.94.61
Kim et al used machine learning algorithms to predict RCC recurrence probabilities within 5- and 10-years post-nephrectomy, utilizing data from 6,849 patients in a Korean RCC database. They developed eight predictive models, achieving an AUC of 0.836 for 5-year recurrence and 0.784 for 10-year recurrence.62 These results highlight the potential of machine learning to support clinical decision-making and enable more personalized post-surgical care.62
Renal clear cell carcinoma (RCC) is a complex disease with unpredictable patient outcomes. While targeted therapies have improved treatment, there remains a need for personalized treatment strategies.67 In this regard Yu et al developed an AI-based predictive model for renal clear cell carcinoma (RCC) using the Unified Perceptual Parsing for Scene Understanding (UPerNet) algorithm to extract CT tumor marker features from a dataset of 267 patients, including 26 treated with targeted drug therapy.63 Patients were categorized into two groups based on survival time (over or under 3 years), and the model achieved high predictive accuracy, with rates of 93.66% for those surviving over 3 years and 94.14% for those under, highlighting the potential of AI to enhance clinical decision-making and support personalized treatment plans to improve patient outcomes and quality of life.63
Testicular Cancer
Testicular cancer is the most prevalent malignancy among adolescent and young adult men in the United States, with approximately 9,000 new cases annually.68
Table 5 outlines the most relevant studies addressing artificial intelligence in testicular cancer, with a focus on study design, AI methodology, and diagnostic or predictive performance.
 |
Table 5 Summary of Principal Articles Regarding AI Applied in Testicular Cancer
|
To date, the application of AI in the management of this oncologic diseases remains limited and still in its early stages, with few studies n conducted on this topic. Among these, the work of Moul et al is particularly noteworthy. Moul and his team aimed to quantify primary tumor histologic factors and incorporate them into a neural network analysis to assess whether staging accuracy could be improved.69 They first constructed a custom AI-model and then compared it with a commercial AI for staging testicular cancer using histologic factors.69 The result obtained showed a superior performance of the custom AI than the commercial AI (92% vs 80% accuracy).69
The lymphovascular invasion is a key prognostic factor in testicular germ cell tumors, particularly in non-seminomatous stage 1 disease.71
The retrospective study of Ghosh et al aimed to develop and evaluate an AI algorithm capable of identifying suspicious areas of lymphovascular invasion in digital whole slide images.70 A total of 184 H&E-stained slides from 19 patients (including seminoma and non-seminomatous cases) were annotated by expert pathologists. A deep learning classifier was then trained for automated segmentation of these regions.70
In a validation set of 118 slides from 10 patients, the model identified 34 suspicious areas, which were reviewed by three expert pathologists to reach a majority consensus. The algorithm achieved a precision of 0.68 for areas flagged as suspicious and 0.56 for confirmed lymphovascular invasion demonstrating the feasibility of an AI-assisted tool that highlights possible lymphovascular invasion for pathologists, who make the final diagnostic decision.70
A recent study led by Lin-Jian Mo et al developed AI-based predictive models (ANN, RF, SVM, and LR) to assess metastasis and treatment response in testicular cancer, using a risk score derived from the expression levels of 12 RNA-binding protein (RBP) genes, including GAPDH, APOBEC3G, ENO1, and HMGA1.72 These models leverage transcriptomic data to predict lymph node metastasis, radiotherapy sensitivity, and chemotherapy efficacy—particularly distinguishing between responses to cisplatin and bleomycin—with the ANN model achieving the highest predictive accuracy.72
Artificial Intelligence Applied for Surgeons and Surgery
Robotic surgery is a key treatment approach for cancers of the urinary system and it offers precious advantages, including lower complication rates for patients compared to traditional open surgery, while enabling surgeons to perform procedures with greater precision.11 Recently, efforts have been made to further enhance robotic surgery by integrating artificial intelligence (AI) to improve its capabilities.73
Augmented reality (AR) refers to systems that overlay digital information onto a real-world view, creating a blended environment. In surgery, this involves projecting preoperative or intraoperative imaging—such as CT, MRI, or ultrasound—onto the operative field, allowing real-time visualization of anatomical structures during laparoscopic or robotic procedures.74 Recent advancements have enabled the integration of AR into urologic procedures, particularly in robot-assisted laparoscopic prostatectomy (RARP) and robotic or laparoscopic partial nephrectomy.75,76 AR has primarily been employed intraoperatively to assist with tumor margin identification and surgical navigation. In addition, AR has been integrated with imaging techniques like SPECT and fluoroscopy to assist in percutaneous renal access and biopsy procedures.77 Although AR-based technology shows considerable potential, current evidence does not clearly demonstrate a significant advantage over conventional techniques. Accurately aligning virtual models with anatomical structures that move or deform during surgery remains a major technical challenge for AR systems.78 For AR to be clinically valuable, its tracking systems must deliver precise, real-time alignment of visual data with the operative field, particularly when dealing with flexible or mobile tissues.79 The most accurate AR solutions often rely on complex multi-camera tracking systems, which substantially increase operational costs and limit widespread adoption.80 The learning curve for radical prostatectomy demonstrates that higher surgical volumes lead to improved patient outcomes s, but relying solely on self-reported caseloads to assess performance is unreliable and inconsistent.12 To address this limitation, various surgical assessment tools have been developed and validated to objectively measure surgical performance such as PACE (Prostatectomy Assessment and Competence Evaluation).81 The PACE tool provides a structured, procedure-specific, and reliable method to objectively assess performance during robot-assisted radical prostatectomy, distinguishing between different expertise levels and offering structured feedback to support tailored training and improve surgical quality.82
The “dVLogger” (Intuitive Surgical) is an innovative data-recording device that captures APMs, including instrument motion tracking metrics and system events in Cartesian coordinates, along with synchronized surgical footage, directly from the da Vinci robotic system in real time during live surgical procedures.83
Hung et al conducted a study to evaluate the effectiveness of machine learning (ML) algorithms in assessing surgical performance throughout the RARP procedure and predicting patient outcomes, while also analyzing the importance of individual automated performance metrics (APMs). They proposed an ML-based method to objectively evaluate surgeons’ performance by analyzing APMs, camera and instrument operation, and energy device usage, offering targeted feedback to both individual surgeons and academic institutions to improve performance evaluation and support efficient, ML-driven, and simulation-based training methods.83
Takeshita et al developed a CNN using a dataset of 1,040 images from intraoperative videos collected across three institutions (2019–2020), split into training and test sets at a 10:3 ratio, to accurately recognize the seminal vesicle and vas deferens (SV-VD) during robot-assisted radical prostatectomy (RARP), demonstrating its potential to enhance SV-VD recognition in posterior RARP, particularly benefiting novice surgeons.84
Natali Rodriguez Peñaranda et al investigated the role of AI in enhancing kidney cancer surgery by addressing challenges in surgical training, highlighting applications such as surgical workflow analysis, instrument annotation, tissue recognition, 3D reconstruction for preoperative planning and intraoperative guidance, and skill evaluation through procedural step identification and instrument tracking.85 While noting that real-time tracking and accurate registration remain obstacles, yet emphasizing AI’s potential to advance training through objective assessments, personalized feedback, and improved learning experiences.85
The role of AI in surgery is still evolving, with research in its early stages. However, progress is expected toward AI-enhanced and AI-assisted techniques, requiring the development of algorithms capable of recognizing and categorizing anatomical features and surgical instruments during operations.86 Despite its growing presence, the use of AI and AR remains a relatively unfamiliar tool for many surgeons, contributing to hesitancy in clinical adoption.87 Reducing the complexity of hardware configuration and eliminating time-consuming manual adjustments are key to making AR accessible for users across all levels of surgical expertise.78
Urologic-Oncology in Patient Care
AI has the potential to enhance patient care by providing clear explanations about surgical procedures, associated risks, and postoperative care.88 Traditionally, patients have relied on search engines like Google for medical information.89
Hershenhouse et al conducted a study in which nine prostate cancer-related questions from Google Trends were classified into diagnosis, treatment, and follow-up. They used ChatGPT-3.5 to generate responses, which were simplified to enhance readability and ensure they were understandable at a sixth-grade reading level suitable for patients.13 ChatGPT-generated responses were considered accurate by 71.7% to 94.3% of urologists and residents.13 These responses were then resubmitted to ChatGPT with a request for further simplified summaries, which were rated accurate in 88.9% of cases and sufficient for decision-making in 88.9% of cases.13 The goal was to create patient-friendly content that remained accurate and informative demonstrating the potential of ChatGPT as a tool for convenient patient education, despite not being explicitly designed for this purpose.13 Additionally, its chatbot interface shows promise for generating accessible and easy-to-read summaries for the general public.88 A key advantage of Large Language Models(LLMs) is their ability to communicate proficiently across various literacy levels and languages, making them powerful tools for engaging marginalized populations.90
However, since accuracy is not flawless, careful selection of information sources is essential to ensure reliability.13
Patients who have access to additional health information resources during the pre- and post-operative phases feel more prepared, more comfortable and experience lower 30-day readmission rates.91
Magdalena Görtz et al developed and evaluated the medical chatbot “prostate cancer communication assistant” (PROSCA), designed to provide information on early prostate cancer detection, prostate diseases, and treatment options.92 Nine men with suspected Prostate Cancer (PCa) tested the chatbot, with 89% reporting increased knowledge, 78% finding it easy to use without assistance and all participants express expressing interest in reusing the chatbot in clinical practice.92 PROSCA proved to be an innovative and effective tool for enhancing doctor-patient communication and improving health education.92
The implementation of machine learning algorithms in clinical practice depends on access to large volumes of patient data, raising significant concerns regarding privacy and data security. Furthermore, the ethical and legal implications of AI-assisted medical decision-making remain complex, with responsibility and accountability for AI-generated outcomes still poorly defined. While many of these challenges may be mitigated over time through technological advancements and improved data governance, further research is essential to ensure safe, transparent, and responsible integration of AI into healthcare.92
These concerns extend to patient-facing applications such as educational chatbots. While tools like ChatGPT and PROSCA can improve accessibility and engagement, they also raise important ethical and regulatory issues. For example, ChatGPT is not designed as a medical device and lacks formal validation processes, which raises concerns about misinformation, over-reliance, and patient autonomy. A global survey by Eppler et al revealed that over 50% of urologists encountered inaccuracies in LLM-generated outputs and most respondents supported the establishment of structured guidelines and oversight mechanisms.93 Similarly, Ganjavi et al reported that only 24% of major publishers provided guidance on generative AI, with no structured policies available for clinical implementation.94 These findings underscore the need for robust validation, ethical review, and regulatory frameworks to ensure responsible deployment of AI tools in patient education.
Although both PROSCA and ChatGPT use natural language processing to facilitate health communication, they differ in development, reliability, and clinical purpose92 PROSCA was designed specifically for prostate cancer education, with medically validated content and structured dialogues, making it suitable for integration into patient counseling pathways.92 ChatGPT, on the other hand, is a general-purpose language model trained on broad internet data without clinical oversight or domain-specific safeguards. Its outputs are generated probabilistically and may vary in accuracy or relevance. While ChatGPT may be useful for improving general health literacy, PROSCA offers a more dependable and clinically appropriate solution for patient-facing applications in urology.92
Implications of ChatGPT in Urologic Medical Writing
Generative AI (GAI) leverages large language models to produce original text or image-based responses to user inputs. Its popularity surged with the introduction of generative pretrained transformers (GPT), particularly with the launch of ChatGPT by OpenAI on November 30, 2022. Within just two months, ChatGPT amassed 100 million monthly users, marking it the fastest-adopted technology in history at that time.
John Mongan’s team introduced CLAIM (Checklist for AI in Medical Imaging) to assist authors and reviewers of AI-related manuscripts in medical imaging.15 This checklist has been expanded to encompass various AI applications, including classification, image reconstruction, text analysis, and workflow optimization,15 and is updated on 2024.95
Eppler et al designed a prospective cross-sectional survey targeting 713 active members of the Young Academic Urologists Urotechnology Group of the European Association of Urology, the uroGPT HUB team (@uroGPT), and additional invited participants. The survey questions were collaboratively developed through an iterative process.93 The primary goal was to assess urologists’ familiarity with ChatGPT, its current applications in urology, potential future uses, ethical challenges, and healthcare regulation.93 In their work approximately 53% of users reported encountering limitations when using ChatGPT or other LLMs in academic settings.93 The most frequently cited challenges included inaccurate responses (44.7%), lack of specificity (42.4%), and inconsistent outputs (26.5%).93 Additionally, there was broad opposition to the use of ChatGPT/LLM outputs in clinical care due to the lack of proper validation.93 Among those utilizing ChatGPT/LLMs in urology, over half (52.4%) acknowledged inaccuracies in the generated outputs, while only 4.9% disagreed.93 Furthermore, the majority of users (59%) indicated they would discontinue using ChatGPT if it were no longer offered for free.93
Moreover, a majority of participants (62.2%) acknowledge the presence of ethical concerns related to the use of ChatGPT in scientific and academic writing such as plagiarism (51.9%), artificial hallucinations (46.1%), and challenges in appropriately citing AI-generated content (30.5%).93 Indeed, nearly 52% of respondents see ChatGPT as a threat to originality, while 48.7% express concerns about its impact on human creativity.93
Ganjavi et al conducted a study focusing on the 100 largest publishers, analyzing their policies to assess the scope and content of the guidance provided to authors regarding the use of generative artificial intelligence (GAI) in academic and scientific journals.94 They noted that only 24% of publishers provided guidelines on the use of GAI in research, with most of these guidelines coming from publishers in the top quartile by journal count. Nearly 90% of the journals currently have such guidelines in place and 19 of them clarified that their GAI guidelines pertain solely to the writing process.94 All publishers explicitly specified that AI cannot be listed as an author.94 In conclusion, they observed a significant variability in the guidance regarding the use of GAI in academic research and scholarly writing reporting that none of the recommendations were developed through a structured, consensus-based guideline process.94
In addition in the work of Eppler et al participants expressed support for establishing clear guidelines (71.5%), for regular monitoring and auditing of AI systems (67.7%), and ethical and legal oversight (65.71%).93
This situation underscores the critical need for the development of unified, cross-disciplinary policies.94
From Hype to Hurdles: Current Limits and Future Paths for AI in Urologic Cancer
Despite the encouraging advances reported across the studies reviewed, several important limitations should be underlined. Much of the current evidence is based on retrospective, single-center, or pilot studies with relatively small patient cohorts, which limits both the external validity and the broader clinical applicability of the findings. Moreover, many of these studies rely exclusively on internal validation methods, without testing their models on independent datasets which is a critical step in demonstrating generalizability. Reported performance metrics such as accuracy, AUC, or precision are often not aligned with clinically meaningful outcomes and are seldom benchmarked against established diagnostic or therapeutic standards. Another key limitation concerns the geographic distribution of the literature: the majority of studies originate from high-income, Western academic institutions, raising questions about their relevance in more diverse or resource-limited settings. All these factors underscore the need for prospective, multicenter trials with standardized protocols and robust external validation to support the safe and effective integration of AI tools into real-world urologic oncology practice.
Conclusion
AI is emerging as a critical tool in urology, offering transformative potential in diagnosis, treatment, and surgical precision. By enhancing imaging analysis, enabling automated pathology workflows, and supporting robotic surgery, AI contributes to better patient outcomes and reduced healthcare burdens. Tools like aSPCNet, DeepSurv, PACE, Paige Prostate demonstrate the ability to improve diagnostic accuracy. Generative AI, including ChatGPT, further expands opportunities for patient education and communication, particularly for marginalized populations.
However, several challenges remain. Ethical concerns around transparency, accountability and bias are yet to be fully addressed. Regulatory frameworks often lag behind technological advancements and many AI models still lack robust clinical validation and external testing across diverse populations. Generalizability remains a key limitation due to variations in data sources, training methodologies and healthcare infrastructures.To allow responsible adoption, future efforts must prioritize data standardization, foster multicenter validation and establish clear ethical and legal guidelines. As AI technology continues to evolve, it will succeed in complement human expertise in urologic oncology provided its integration is guided by scientific rigor, inclusivity, and patient-centered principles.
Future studies should focus on real-world validation, cross-institutional collaborations, and rigorous evaluation of AI’s clinical impact. As the technology continues to evolve, AI is poised to bridge the gap between human expertise and machine intelligence, reshaping urologic oncology toward more personalized, efficient, and equitable care.
Acknowledgments
The authors wish to acknowledge the contributions of colleagues and peers who provided insightful feedback during the development of this review. No external funding was received for the preparation of this manuscript.
Disclosure
The authors report no conflicts of interest in this work.
References
1. Checcucci E, Autorino R, Cacciamani GE, et al. Artificial intelligence and neural networks in urology: current clinical applications. Minerva Urol E Nefrol Ital J Urol Nephrol. 2020;72(1). doi:10.23736/S0393-2249.19.03613-0
2. Hodges A. Computer science. Beyond Turing’s machines. Science. 2012;336(6078):163–164. doi:10.1126/science.1218417
3. Darcy AM, Louie AK, Roberts LW. Machine learning and the profession of medicine. JAMA. 2016;315(6):551–552. doi:10.1001/jama.2015.18421
4. Bajwa J, Munir U, Nori A, Williams B. Artificial intelligence in healthcare: transforming the practice of medicine. Future Healthc J. 2021;8(2):e188–e194. doi:10.7861/fhj.2021-0095
5. Shah M, Naik N, Somani BK, Hameed BZ. Artificial intelligence (AI) in urology-Current use and future directions: an iTRUE study. Turk J Urol. 2020;46(Suppl 1):S27–S39. doi:10.5152/tud.2020.20117
6. The role of artificial intelligence in urology. Hellenic Urology. 2023;35(1):16–19.
7. Gómez Rivas J, Toribio Vázquez C, Ballesteros Ruiz C, et al. Artificial intelligence and simulation in urology. Actas Urol Esp Engl Ed. 2021;45(8):524–529. doi:10.1016/j.acuroe.2021.07.001
8. Barragán-Montero A, Javaid U, Valdés G, et al. Artificial intelligence and machine learning for medical imaging: a technology review. Phys Medica PM Int J Devoted Appl Phys Med Biol off J Ital Assoc Biomed Phys AIFB. 2021;83:242–256. doi:10.1016/j.ejmp.2021.04.016
9. Kriegeskorte N, Golan T. Neural network models and deep learning. Curr Biol CB. 2019;29(7):R231–R236. doi:10.1016/j.cub.2019.02.034
10. Qin S, Chislett B, Ischia J, et al. ChatGPT and generative AI in urology and surgery—A narrative review. BJUI Compass. 2024;5(9):813–821. doi:10.1002/bco2.390
11. Ferretti S, Dell’Oglio P, Ciavarella D, Galfano A, Schips L, Marchioni M. Retzius-sparing robotic-assisted prostatectomy: technical challenges for surgeons and key prospective refinements. Res Rep Urol. 2023;15:541–552. doi:10.2147/RRU.S372803
12. Abboudi H, Khan MS, Guru KA, et al. Learning curves for urological procedures: a systematic review. BJU Int. 2014;114(4):617–629. doi:10.1111/bju.12315
13. Hershenhouse JS, Mokhtar D, Eppler MB, et al. Accuracy, readability, and understandability of large language models for prostate cancer information to the public. Prostate Cancer Prostatic Dis. 2024:1–6. doi:10.1038/s41391-024-00826-y
14. Chatrian A, Colling RT, Browning L, et al. Artificial intelligence for advance requesting of immunohistochemistry in diagnostically uncertain prostate biopsies. Mod Pathol off J U S Can Acad Pathol Inc. 2021;34(9):1780–1794. doi:10.1038/s41379-021-00826-6
15. Mongan J, Moy L, Kahn CE. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): a guide for authors and reviewers. Radiol Artif Intell. 2020;2(2):e200029. doi:10.1148/ryai.2020200029
16. Cancer Today. Available from: https://gco.iarc.who.int/today/.
17. AIOM genitourinary malignancies. AIOM; 2023. Available from: https://www.aiom.it/tumore-della-prostata-nel-2022-40-500-diagnosi-16-in-cinque-anni-aiom-dallintelligenza-artificiale-la-nuova-via-per-scegliere-le-cure/.
18. Venkatramani V. Urovision 2020: the future of urology. Indian J Urol IJU J Urol Soc India. 2015;31(2):150–155. doi:10.4103/0970-1591.152918
19. Checcucci E, De Cillis S, Granato S, et al. Applications of neural networks in urology: a systematic review. Curr Opin Urol. 2020;30(6):788–807. doi:10.1097/MOU.0000000000000814
20. Siegel RL, Kratzer TB, Giaquinto AN, Sung H, Jemal A. Cancer statistics, 2025. CA Cancer J Clin. 2025;75(1):10–45. doi:10.3322/caac.21871
21. Seetharaman A, Bhattacharya I, Chen LC, et al. Automated detection of aggressive and indolent prostate cancer on magnetic resonance imaging. Med Phys. 2021;48(6):2960–2972. doi:10.1002/mp.14855
22. Saha A, Bosma JS, Twilt JJ, et al. Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study. Lancet Oncol. 2024;25(7):879–887. doi:10.1016/S1470-2045(24)00220-1
23. Islam R, Imran A, Rabbi MF. Prostate cancer detection from MRI using efficient feature extraction with transfer learning. Prostate Cancer. 2024;2024:1588891. doi:10.1155/2024/1588891
24. Bulten W, Kartasalo K, Chen PHC, et al. Artificial intelligence for diagnosis and Gleason grading of prostate cancer: the PANDA challenge. Nat Med. 2022;28(1):154–163. doi:10.1038/s41591-021-01620-2
25. Bianchi G, Puliatti S, Rodriguez Peñaranda N. Artificial intelligence evaluation of confocal microscope prostate images: our preliminary experience. Minerva Urology and Nephrology. 2023;75(5):545–547. doi:10.23736/S2724-6051.23.05538-6
26. Rana A, Lowe A, Lithgow M, et al. Use of deep learning to develop and analyze computational hematoxylin and eosin staining of prostate core biopsy images for tumor diagnosis. JAMA Netw Open. 2020;3(5):e205111. doi:10.1001/jamanetworkopen.2020.5111
27. Tolkach Y, Ovtcharov V, Pryalukhin A, et al. An international multi-institutional validation study of the algorithm for prostate cancer detection and Gleason grading. NPJ Precis Oncol. 2023;7(1):77. doi:10.1038/s41698-023-00424-6
28. Zhu L, Pan J, Mou W, et al. Harnessing artificial intelligence for prostate cancer management. Cell Rep Med. 2024;5(4):101506. doi:10.1016/j.xcrm.2024.101506
29. Hung AJ, Chen J, Ghodoussipour S, et al. Deep learning on automated performance metrics and clinical features to predict urinary continence recovery after robot-assisted radical prostatectomy. BJU Int. 2019;124(3):487–495. doi:10.1111/bju.14735
30. Orsini A, Ferretti S, Porreca A, et al. PI-RADS in predicting csPCa: a comparison between academic and nonacademic centers. The Prostate. 2025;85(4):337–343. doi:10.1002/pros.24832
31. Epstein JI, Egevad L, Amin MB, et al. The 2014 International Society of Urological Pathology (ISUP) consensus conference on Gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system. Am J Surg Pathol. 2016;40(2):244. doi:10.1097/PAS.0000000000000530
32. Ström P, Kartasalo K, Olsson H, et al. Artificial intelligence for diagnosis and grading of prostate cancer in biopsies: a population-based, diagnostic study. Lancet Oncol. 2020;21(2):222–232. doi:10.1016/S1470-2045(19)30738-7
33. Jung M, Jin MS, Kim C, et al. Artificial intelligence system shows performance at the level of uropathologists for the detection and grading of prostate cancer in core needle biopsy: an independent external validation study. Mod Pathol off J U S Can Acad Pathol Inc. 2022;35(10):1449–1457. doi:10.1038/s41379-022-01077-9
34. Ryu HS, Jin MS, Park JH, et al. Automated Gleason scoring and tumor quantification in prostate core needle biopsy images using deep neural networks and its comparison with pathologist-based assessment. Cancers. 2019;11(12):1860. doi:10.3390/cancers11121860
35. Steiner DF, Nagpal K, Sayres R, et al. Evaluation of the use of combined artificial intelligence and pathologist assessment to review and grade prostate biopsies. JAMA Netw Open. 2020;3(11):e2023267. doi:10.1001/jamanetworkopen.2020.23267
36. Kartasalo K, Ström P, Ruusuvuori P, et al. Detection of perineural invasion in prostate needle biopsies with deep neural networks. Virchows Arch Int J Pathol. 2022;481(1):73–82. doi:10.1007/s00428-022-03326-3
37. Harmon SA, Patel PG, Sanford TH, et al. High throughput assessment of biomarkers in tissue microarrays using artificial intelligence: PTEN loss as a proof-of-principle in multi-center prostate cancer cohorts. Mod Pathol off J U S Can Acad Pathol Inc. 2021;34(2):478–489. doi:10.1038/s41379-020-00674-w
38. Ambrosini P, Hollemans E, Kweldam CF, Van leenders GJLH, Stallinga S, Vos F. Automated detection of cribriform growth patterns in prostate histology images. Sci Rep. 2020;10(1):14904. doi:10.1038/s41598-020-71942-7
39. Bashashati A, Goldenberg SL. AI for prostate cancer diagnosis – hype or today’s reality? Nat Rev Urol. 2022;19(5):261–262. doi:10.1038/s41585-022-00583-4
40. Haghighat M, Browning L, Sirinukunwattana K, et al. Automated quality assessment of large digitised histology cohorts by artificial intelligence. Sci Rep. 2022;12(1):5002. doi:10.1038/s41598-022-08351-5
41. Hamdy FC, Donovan JL, Lane JA, et al. Fifteen-year outcomes after monitoring, surgery, or radiotherapy for prostate cancer. N Engl J Med. 2023;388(17):1547–1558. doi:10.1056/NEJMoa2214122
42. Marchioni M, Primiceri G, Castellan P, et al. Conservative management of urinary incontinence following robot-assisted radical prostatectomy. Minerva Urol E Nefrol Ital J Urol Nephrol. 2020;72(5):555–562. doi:10.23736/S0393-2249.20.03782-0
43. Marchioni M, De Francesco P, Castellucci R. Management of erectile dysfunction following robot-assisted radical prostatectomy: a systematic review. Minerva Urologica e Nefrologica. 2020;72(5):543–554. doi:10.23736/S0393-2249.20.03780-7
44. International Agency for Research on Cancer. Estimated number of new cases in 2020, worldwide, both sexes, all ages. World Health Organization; 2021. Available from: https://gco.iarc.who.int/today/.
45. De Nunzio C, Giannatempo P, Passalacqua R, Fiorini E, Luccarini I, Brigido A. Epidemiology and unmet needs of bladder cancer in Italy: a critical review. Minerva Urol E Nefrol Ital J Urol Nephrol. 2020;72(1):1–12. doi:10.23736/S0393-2249.19.03498-2
46. Xu X, Zhang X, Tian Q, et al. Three-dimensional texture features from intensity and high-order derivative maps for the discrimination between bladder tumors and wall tissues via MRI. Int J Comput Assist Radiol Surg. 2017;12(4):645–656. doi:10.1007/s11548-017-1522-8
47. Garapati SS, Hadjiiski L, Cha KH, et al. Urinary bladder cancer staging in CT urography using machine learning. Med Phys. 2017;44(11):5814–5823. doi:10.1002/mp.12510
48. Ikeda A, Nosato H, Kochi Y, et al. Support system of cystoscopic diagnosis for bladder cancer based on artificial intelligence. J Endourol. 2020;34(3):352–358. doi:10.1089/end.2019.0509
49. Fahoum I, Naamneh R, Silberberg K, Hagege R, Hershkovitz D. Detection of muscularis propria invasion in urothelial carcinoma using artificial intelligence. Technol Cancer Res Treat. 2024;23:15330338241257479. doi:10.1177/15330338241257479
50. He C, Xu H, Yuan E, et al. The accuracy and quality of image-based artificial intelligence for muscle-invasive bladder cancer prediction. Insights Imaging. 2024;15(1):185. doi:10.1186/s13244-024-01780-y
51. Suartz CV, Martinez LM, Cordeiro MD, et al. Artificial intelligence for predicting response to neoadjuvant chemotherapy for bladder cancer A comprehensive systematic review and meta-analysis. Can Urol Assoc J J Assoc Urol Can. 2024;18(9):E276–E284. doi:10.5489/cuaj.8681
52. Lotan Y, Krishna V, Abuzeid WM, et al. Predicting response to intravesical bacillus Calmette-Guérin in high-risk nonmuscle-invasive bladder cancer using an artificial intelligence-powered pathology assay: development and validation in an International 12-center cohort. J Urol. 2025;213(2):192–204. doi:10.1097/JU.0000000000004278
53. Soria F, Shariat SF, Lerner SP, et al. Epidemiology, diagnosis, preoperative evaluation and prognostic assessment of upper-tract urothelial carcinoma (UTUC). World J Urol. 2017;35(3):379–387. doi:10.1007/s00345-016-1928-x
54. Salminen AP, Jambor I, Syvanen KT, Bostrom PJ. Update on novel imaging techniques for the detection of lymph node metastases in bladder cancer. Minerva Urol E Nefrol Ital J Urol Nephrol. 2016;68(2):138–149.
55. Wu S, Chen X, Pan J, et al. An artificial intelligence system for the detection of bladder cancer via cystoscopy: a multicenter diagnostic study. J Natl Cancer Inst. 2022;114(2):220–227. doi:10.1093/jnci/djab179
56. Ferro M, Falagario UG, Barone B, et al. Artificial intelligence in the advanced diagnosis of bladder cancer-comprehensive literature review and future advancement. Diagnostics. 2023;13(13):2308. doi:10.3390/diagnostics13132308
57. Liow E, Tran B. Precision oncology in urothelial cancer. ESMO Open. 2020;5(Suppl 1):e000616. doi:10.1136/esmoopen-2019-000616
58. Kujtan L, Hussain A, Subramanian J, Masood A. The evolving genomic landscape in urothelial cancer. Curr Opin Oncol. 2018;30(3):197–202. doi:10.1097/CCO.0000000000000447
59. Cicchetti R, Basconi M, Litterio G, et al. Advances in molecular mechanisms of kidney disease: integrating renal tumorigenesis of hereditary cancer syndrome. Int J Mol Sci. 2024;25(16):9060. doi:10.3390/ijms25169060
60. Fenstermaker M, Tomlins SA, Singh K, Wiens J, Morgan TM. Development and validation of a deep-learning model to assist with renal cell carcinoma histopathologic interpretation. Urology. 2020;144:152–157. doi:10.1016/j.urology.2020.05.094
61. Choi JW, Hu R, Zhao Y, et al. Preoperative prediction of the stage, size, grade, and necrosis score in clear cell renal cell carcinoma using MRI-based radiomics. Abdom Radiol N Y. 2021;46(6):2656–2664. doi:10.1007/s00261-020-02876-x
62. Kim H, Lee SJ, Park SJ, Choi IY, Hong SH. Machine learning approach to predict the probability of recurrence of renal cell carcinoma after surgery: prediction model development study. JMIR Med Inform. 2021;9(3):e25635. doi:10.2196/25635
63. Yu Y, Niu J, Yu Y, Xia S, Sun S. AI predictive modeling of survival outcomes for renal cancer patients undergoing targeted therapy. Sci Rep. 2024;14(1):26156. doi:10.1038/s41598-024-77638-6
64. Yang R, Wu J, Sun L, et al. Radiomics of small renal masses on multiphasic CT: accuracy of machine learning-based classification models for the differentiation of renal cell carcinoma and angiomyolipoma without visible fat. Eur Radiol. 2020;30(2):1254–1263. doi:10.1007/s00330-019-06384-5
65. Khene ZE, Kammerer-Jacquet SF, Bigot P, et al. Clinical application of digital and computational pathology in renal cell carcinoma: a systematic review. Eur Urol Oncol. 2024;7(3):401–411. doi:10.1016/j.euo.2023.10.018
66. Distante A, Marandino L, Bertolo R, et al. Artificial intelligence in renal cell carcinoma histopathology: current applications and future perspectives. Diagnostics. 2023;13(13):2294. doi:10.3390/diagnostics13132294
67. Marchioni M, Bandini M, Preisser F, et al. Survival after cytoreductive nephrectomy in metastatic non-clear cell renal cell carcinoma patients: a population-based study. Eur Urol Focus. 2019;5(3):488–496. doi:10.1016/j.euf.2017.11.012
68. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7–30. doi:10.3322/caac.21590
69. Moul JW, Snow PB, Fernandez EB, Maher PD, Sesterhenn IA. Neural network analysis of quantitative histological factors to predict pathological stage in clinical stage I nonseminomatous testicular cancer. J Urol. 1995;153(5):1674–1677. doi:10.1016/S0022-5347(01)67502-5
70. Ghosh A, Sirinukunwattana K, Khalid Alham N, et al. The potential of artificial intelligence to detect lymphovascular invasion in testicular cancer. Cancers. 2021;13(6):1325. doi:10.3390/cancers13061325
71. International Germ Cell Cancer Collaborative Group. International germ cell consensus classification: a prognostic factor-based staging system for metastatic germ cell cancers. J Clin Oncol off J Am Soc Clin Oncol. 1997;15(2):594–603. doi:10.1200/JCO.1997.15.2.594
72. Mo LJ, Liang HQ, Yu ZY, et al. RNA-binding protein expression based machine learning model predicts metastasis and treatment outcome of testicular cancer. Genes Genomics. 2025;47(6):741–759. doi:10.1007/s13258-025-01636-9
73. Tamborino F, Stritto GD, Salzano G, et al. Current status, evolution, and future perspectives in robotic platform systems for prostate cancer treatment: a narrative review. Chin Clin Oncol. 2024;13(5):74. doi:10.21037/cco-24-47
74. Hamacher A, Kim SJ, Cho ST, et al. Application of virtual, augmented, and mixed reality to urology. Int Neurourol J. 2016;20(3):172–181. doi:10.5213/inj.1632714.357
75. Porpiglia F, Checcucci E, Amparore D, et al. Three-dimensional augmented reality robot-assisted partial nephrectomy in case of complex tumours (Padua ≥10): a new intraoperative tool overcoming the ultrasound guidance. Eur Urol. 2020;78(2):229–238. doi:10.1016/j.eururo.2019.11.024
76. Porpiglia F, Checcucci E, Amparore D, et al. Augmented-reality robot-assisted radical prostatectomy using hyper-accuracy three-dimensional reconstruction (HA3DTM) technology: a radiological and pathological study. BJU Int. 2019;123(5):834–845. doi:10.1111/bju.14549
77. Mu Y, Hocking D, Wang ZT, Garvin GJ, Eagleson R, Peters TM. Augmented reality simulator for ultrasound-guided percutaneous renal access. Int J Comput Assist Radiol Surg. 2020;15(5):749–757. doi:10.1007/s11548-020-02142-x
78. Reis G, Yilmaz M, Rambach J, et al. Mixed reality applications in urology: requirements and future potential. Ann Med Surg 2012. 2021;66:102394. doi:10.1016/j.amsu.2021.102394
79. Bertolo R, Hung A, Porpiglia F, Bove P, Schleicher M, Dasgupta P. Systematic review of augmented reality in urological interventions: the evidences of an impact on surgical outcomes are yet to come. World J Urol. 2020;38(9):2167–2176. doi:10.1007/s00345-019-02711-z
80. Caruso D, Eudes A, Sanfourche M, Vissière D, Le Besnerais G. A robust indoor/outdoor navigation filter fusing data from vision and magneto-inertial measurement unit. Sensors. 2017;17(12):2795. doi:10.3390/s17122795
81. Goh AC, Goldfarb DW, Sander JC, Miles BJ, Dunkin BJ. Global evaluative assessment of robotic skills: validation of a clinical assessment tool to measure robotic surgical skills. J Urol. 2012;187(1):247–252. doi:10.1016/j.juro.2011.09.032
82. Hussein AA, Ghani KR, Peabody J, et al. Development and validation of an objective scoring tool for robot-assisted radical prostatectomy: prostatectomy assessment and competency evaluation. J Urol. 2017;197:1237–1244. doi:10.1016/j.juro.2016.11.100
83. Hung AJ, Chen J, Che Z, et al. Utilizing machine learning and automated performance metrics to evaluate robot-assisted radical prostatectomy performance and predict outcomes. J Endourol. 2018;32(5):438–444. doi:10.1089/end.2018.0035
84. Takeshita N, Sakamoto S, Kitaguchi D, et al. Deep learning-based seminal vesicle and vas deferens recognition in the posterior approach of robot-assisted radical prostatectomy. Urology. 2023;173:98–103. doi:10.1016/j.urology.2022.12.006
85. Rodriguez Peñaranda N, Eissa A, Ferretti S, et al. Artificial intelligence in surgical training for kidney cancer: a systematic review of the literature. Diagn Basel Switz. 2023;13(19):3070. doi:10.3390/diagnostics13193070
86. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019;25(1):44–56. doi:10.1038/s41591-018-0300-7
87. Amparore D, Piramide F, Verri P, et al. New generation of 3D virtual models with perfusional zones: perioperative assistance for the best pedicle management during robotic partial nephrectomy. Curr Oncol. 2023;30(4):4021–4032. doi:10.3390/curroncol30040304
88. Nedbal C, Naik N, Castellani D, Gauhar V, Geraghty R, Somani BK. ChatGPT in urology practice: revolutionizing efficiency and patient care with generative artificial intelligence. Curr Opin Urol. 2024;34(2):98–104. doi:10.1097/MOU.0000000000001151
89. Checcucci E, Rodler S, Piazza P, Porpiglia F, Cacciamani GE. Transitioning from “Dr. Google” to “Dr. ChatGPT”: the advent of artificial intelligence chatbots. Transl Androl Urol. 2024;13(6):1067–1070. doi:10.21037/tau-23-629
90. Rodler S, Ganjavi C, Backer PD, et al. Generative artificial intelligence in surgery. Surgery. 2024;175(6):1496–1502. doi:10.1016/j.surg.2024.02.019
91. Martin LA, Finlayson SRG, Brooke BS. Patient preparation for transitions of surgical care: is failing to prepare surgical patients preparing them to fail? World J Surg. 2017;41(6):1. doi:10.1007/s00268-017-3884-z
92. Görtz M, Baumgärtner K, Schmid T, et al. An artificial intelligence-based chatbot for prostate cancer education: design and patient evaluation study. Digit Health. 2023;9:20552076231173304. doi:10.1177/20552076231173304
93. Eppler M, Ganjavi C, Ramacciotti LS, et al. Awareness and use of ChatGPT and large language models: a prospective cross-sectional global survey in urology. Eur Urol. 2024;85(2):146–153. doi:10.1016/j.eururo.2023.10.014
94. Ganjavi C, Eppler MB, Pekcan A, et al. Publishers’ and journals’ instructions to authors on use of generative artificial intelligence in academic and scientific publishing: bibliometric analysis. THE BMJ. 2024;384:e077192. doi:10.1136/bmj-2023-077192
95. Tejani AS, Klontzas ME, Gatti AA, et al. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): 2024 Update. Radiol Artif Intell. 2024;6(4):e240300. doi:10.1148/ryai.240300