Artificial intelligence

AI is commonly defined as the set of computational technologies that emulate intellectual processes typically associated with human cognition, such as reasoning, learning, and problem solving [2]. Human intelligence can be metaphorically described as the union of "software" with its "hardware"; the "software" includes the algorithms and knowledge we use to process information, which can be enhanced throughout life. Conversely, the "hardware" represents the physical structure of the human brain, which is less susceptible to rapid changes compared to software, due to the gradual pace of biological evolution over generations. Interestingly, AI is not restricted by such evolutionary constraints. It has the unique capability to upgrade both its "software" and "hardware" without the need for generational evolution. This reflects the progression of AI development, which has advanced from merely replicating basic human cognitive functions to enhancing and expediting these processes while adeptly handling extensive data, culminating in the creation of sophisticated, advanced systems. When integrated into healthcare, AI aligns with the core principles of scientific research, which involve the meticulous processing of clinical data—the "input"—and the thorough examination of clinical observations—the "output"— to establish and validate interconnected patterns. There are several subfields of AI [3, 4]:
● Machine learning (ML) integrates computer science, statistics, and mathematics to develop algorithms capable of predicting outcomes. These algorithms excel over traditional statistics by focusing on predictions rather than relationships between variables and continuously improve by learning from additional data.
● Deep learning (DL), an advanced branch of ML, utilizes artificial neural networks to analyze data with a complexity that resembles the human brain. DL systems identify patterns and features within medical imagery, often discovering diagnostic indicators that human analysis might miss.
● Big data encompasses vast data sets that traditional software cannot handle. In healthcare, AI processes this data to identify disease patterns, predict disease progression, and aid in drug development. The unlimited potential offered by the evolution of these models has, over the years, found extensive applications in the field of urology, ranging from functional urology to uro-oncology.
In the realm of urolithiasis, the main applications of AI have varied from outcomes prediction to diagnosis and therapy [5]. The prediction of outcomes is the most extensively studied area; stone-free status, the detection of infection, the optimization of kidney stone fragmentation, and the prediction of stone patients' health-related quality of life are some of the outcomes investigated. Predictive applications extend to the differential diagnosis and even to predictions of stone composition. Advances have also been made in the surgical field, such as improving the surgical performance by refining protocols, stone localizations, and patient selection [6, 7]. Urogynecology has also adopted ML systems, particularly for assessing functional outcomes and postoperative results [8]. Predictive models have been employed to evaluate the occurrence of complications such as stress urinary incontinence following prolapse surgery, conditions recurrence, and overall outcomes of surgical interventions. Among benign urological conditions, the heterogenicity of benign prostatic hyperplasia (BPH) has provided a broad scope for the integration of AI. Computer vision-based systems have been evaluated for their diagnostic accuracy in identifying BPH histologically, with results showing a high accuracy level. Neural networks have been applied to support the prediction of complications following BPH-related surgeries, to examine the risk of worsening symptoms, and to analyze the contributing risk factors. Additionally, these networks have been instrumental in developing models to predict patient responses to medical treatments [6]. Urological cancer has reaped the greatest benefits from the development of AI. More data, more connections, and generally more inputs allow the potential of machine learning and its derivatives to be fully realized. Hence, it is clear how urological cancer provides an immense array of factors that are well-suited to the advantages of these new technologies, ranging from pathogenesis and risk factors to diagnosis, treatment decisions, and outcome predictions [9].
In the diagnostic field, major advancements have been achieved in the management of Prostate cancer (PCa), including the use of clinico-pathological data to differentiate between organ-confined and non-organ-confined PCa [10, 11]. Algorithms have been developed to predict Gleason Scores based on MRI studies or to detect PCa on digitized pathology images. Similar concepts have been applied to urothelial and renal cell carcinoma (RCC). Diagnosis and staging of urothelial cancer now potentially rely on imaging and biomarkers, while metabolomic data assist in differentiating between RCC and healthy tissue [12]. These concepts have also been applied to the prediction of oncological outcomes such as disease recurrence, survival analysis, and the guidance of management decision-making and therapy selection [13]. The continual advancement of AI in urology signifies a transformative shift towards data-driven precision medicine (Figure 1). By harnessing sophisticated algorithms and computational analytics, we are not only redefining the paradigms of diagnosis and treatment but also empowering clinicians with predictive insights that could lead to more effective, individualized patient care strategies, ultimately enhancing outcomes and elevating the standard of urological practice.

Augmented reality

AI algorithms have also a critical role in extended reality (XR) technologies implementation, that provides clinicians with data-driven insights for decision-making and real-time precision-guided interventions.
XR encompasses a spectrum of immersive technologies that alter and enhance the traditional interface between the user and the computer, each with distinct applications in urology.
● Virtual reality (VR) immerses users completely within a digital environment, isolating them from the physical world.
● Augmented reality (AR) on the other hand, superimposes digital images onto the user's real-world view, typically through digital interfaces that allows for clear visibility of the environment.
● Mixed reality (MR) combines elements of both AR and VR, merging real and virtual worlds to produce new environments where physical and digital objects co-exist and interact in real time.
XR's role in urology extends to education and skill development, where devices like VR headsets and optical headmounted displays (OHMDs) create immersive learning scenarios [14]. VR has progressed from facilitating the practice of fundamental, discrete skills to enabling comprehensive procedural simulations. Likewise, AR offers to urology trainees the opportunity to engage with highly detailed, interactive models, thereby enriching their learning experience and surgical training. The validated fields of application range from uro-oncology to endourology and andrology; however, there is a lack of validated VR simulators for the management of urological emergencies [15].
Despite issues with the quality of studies available and the cost-effectiveness of these advanced training devices compared to the "standard" ones, the early results are largely in favor of the integration of XR simulations in surgical training [16]. Recent advances have led to an even more revolutionary application of AR and MR in the field of Urology. As a matter of fact, the integration of these technologies is actively being explored as a potential game-changer for urologic surgeries [17]. A growing body of evidence suggests that AR and MR can significantly enhance the interactivity of preoperative planning and patient education by integrating standard imaging with patient-specific 3D models [18]. These models can be either printed, visualized through AR, or displayed on 3D/2D computer monitors, offering a more comprehensive and tailored surgical planning and educational process for both surgeons and patients alike. An area where this technology has found an ideal application is in nephron-sparing renal surgery, where the integration of advanced information about tumor position and vascularization and relations with vital structures has been shown to enhance preoperative planning and surgeon's confidence, as well as the patient's understanding of their condition [19]. Moreover, AR can be applied during a surgical intervention to permit augmented visualization directly within the surgical field, thereby integrating seamlessly with the clinician's workflow. By enabling the superimposition of diagnostic preoperative images onto the operative field, AR offers enhanced visual guidance for identification of anatomical structures during complex procedures. Additionally, it can assist in minimizing the risk of complications by indicating the real-time location of fragile structures [20]. Once again, a highly researched field of application has been that of partial nephrectomies, but various uro-oncological and reconstructive procedures have also been evaluated [21].

Challenges and ethical concerns

As we embrace the integration of AI and AR in urology, we confront several challenges and ethical dilemmas that must be addressed to ensure responsible adoption and application of these technologies [22, 23]. One of the primary challenges is data privacy. As healthcare increasingly relies on digital data, ensuring the confidentiality and security of patient information is paramount. AI systems require vast datasets for training and validation, which raises concerns about the potential for data breaches and misuse. Potential biases in AI algorithms also pose significant challenges. AI models can inadvertently perpetuate existing biases present in their training data, leading to disparities in patient care. This is particularly concerning in medicine, where algorithms may influence decisionmaking in patient diagnosis and treatment. The implications of automation on clinical decision-making are also a critical consideration. There is a delicate balance to be struck between leveraging AI for more accurate diagnoses and treatment plans and ensuring that these technologies support, rather than overshadow, the clinician's expertise. Furthermore, as AI and AR tools become more prevalent, the medical community must grapple with their impact on the clinician-patient relationship. There is a risk that over-reliance on technology could depersonalize care and diminish the role of the clinician's judgment, experience, and personal interaction with the patient.