Bladder cancer (BCa) is a common urological malignancy of the uroepithelium. As the tenth most common cancer in the world, BCa accounts for over 573,000 new cases and 213,000 deaths annually, with a higher incidence in men compared to women (1). According to the latest cancer statistics from the American Cancer Society, BCa has become the fourth most common malignancy in the American male population (2). Based on the above facts, it is of great significance to achieve early diagnosis and treatment of BCa.

The diagnostic methods for BCa include imaging, cystoscopy, urine tests, etc. and pathological diagnosis is the gold standard (3, 4). BCa can be divided into two types in pathology according to the degree of muscle invasion: non-muscle-invasive BCa and muscle-invasive BCa (5), and prognosis of them is very different. It requires pathologists’ efforts to identify them, but the manual review of pathologist may sometimes bring mistakes (4). Imaging tests can help detect BCa early (6), but they are also influenced by human factors. Treatments for BCa include non-surgical treatments like systemic chemotherapy and radiation therapy and surgical treatments like transurethral resection of bladder tumor, open radical cystectomy, laparoscopic radical cystectomy and robotic-assisted laparoscopic radical cystectomy. Doctors play an important role in treatment selection and prognostic prediction, but this can be influenced by subjective factors (7, 8). Robotic surgery has been widely used in the treatment of BCa, but it is highly dependent on the control of doctors and has not yet been automated. In summary, diagnosis and treatment of BCa can still be improved, and artificial intelligence (AI) brings hope for it.

AI was first proposed in 1956 at Dartmouth Conferences (9) which is the science of research and development for the simulation, extension and expansion of human intelligence. Due to its powerful learning and computational capabilities, AI has already been used in the field of BCa throughout the entire process of medical treatment. In recent years, many AI-based pathological diagnostic models for BCa have been designed (10). A convolutional neural network algorithm-based urinary cytology diagnostic model was shown to be effective in reducing the underdiagnosis rate of low-grade uroepithelial carcinoma (11). The robotic surgery, which has been widely applied in the treatment of BCa in recent years, is also an important area for AI applications. Some studies have reported the use of machine learning to assess surgeons’ robotic surgical skills and stylistic behavior (12, 13). An attention-guided network for surgical instrument segmentation from endoscopic images was invented to improve the accuracy of robotic surgery (14). Meanwhile, the application of AI opens up the possibility of building autonomous surgical robots (15). A study demonstrated a new framework for planning and executing semi-autonomous tissue retraction in robotic surgery based on deep learning and procedural algorithms (16). In addition, AI can also be used to predict the prognosis of treatment, such as recurrence risk and survival prediction (17).

Unfortunately, there are only a few reviews, systematic reviews and meta-analyses that provide a comprehensive analysis of the application of AI in the field of BCa, and only a relatively small number of them have addressed the focus and frontiers. As a visual analysis method, bibliometric analysis has been widely used in different research fields to analyze hotspots (18–26). Its principle is to quantitatively analyze the literature data in a certain research area based on the knowledge of mathematics, statistics, computer science and other disciplines (23, 24). Specifically, bibliometric analysis is used to analyze publications to obtain information about influence in countries/regions, institutions, journals, authors, references, keywords, etc. (25). In addition, it is also used to obtain information about cooperation and relationships by mapping social networks (25). Consequently, to provide an in-depth and reliable analysis of AI application in BCa, bibliometric analysis and visualization of the relevant literature from 2000 to 2022 were performed. We hope this study will help researchers understand the overall framework of the field and today’s research hotspots.

Researchers need to update their professional knowledge by reading a large number of documents. Bibliometric analysis helps to clearly show the research progress and hotspots in relevant fields.

Breakthroughs in machine learning algorithms were made in 2013, leading to the rapid expansion of AI into almost all fields. By 2016, AI had begun to cover many medical areas (42). Supportive policies from governments and funding agencies have also provided funding for the growth in the field of AI application in BCa (42). In 2019, the global publication volume in this field experienced a surge, with an impressive growth rate of 43.98%. This field continues to rapidly increase and is considered a research area with great potential.

In terms of countries/regions, the country/region involved in research about AI application in BCa are mainly concentrated in North America, East Asia, and Europe. These regions are the most economically and technologically advanced regions in the world. But the development level and scale of research about AI application in BCa in developing countries are not high, which indicates that research needs sufficient funding and mature science and technology to support it. In addition, China is the only developing country in the top 10. Notably, China surpassed the US in the annual publishing volume in 2021 and 2022, and has maintained a high rate of increase. It can be expected that China will invest more money and effort in this field and may achieve more influence in the future.

Regarding country/region cooperation, cooperation between countries/regions occurs mainly between developed countries in North America and Europe. Specifically, the U.S. is at the center of global collaboration, with significant cooperation with China, Italy, Canada, the UK, and Germany. In addition, Germany, Italy, and the UK also have very high TLS in co-authorship analysis, indicating their essential role in global collaboration. This collaboration facilitates the output of high-level research and the explosion of progress in cutting-edge science and technology. However, it can also lead to a high scientific and technological disparity between countries, resulting in more severe scientific and technical polarization. To solve this, developing countries should proactively seek international cooperation.

Of the top 10 research institutions, eight are from the United States, one is from Austria, and one is from Sweden. With the highest number of publications (56), the highest citation rate (5045), and the highest per-page citation rate (90.09), the University of North Carolina is considered the most influential research institution in the field of AI application in BCa. The University of North Carolina has been working in this field since 1999, focusing on robotic-assisted radical cystectomy, robotic-assisted urinary diversion surgery, BCa prediction models, and AI algorithms for radionics. Roswell Park Comprehensive Cancer Center has the highest TLS (5319). Although Roswell Park Comprehensive Cancer Center entered this field later (2007), it achieved high-quality research (average citations per paper = 58.63) with a research focus similar to the University of North Carolina. Two European institutions are in the top 10 worldwide in terms of publication volume. The average citations per paper were 19.81 for Medical University of Vienna and 31.19 for Karolinska Institutet, which still have a gap with the top institutions in the U.S. Currently, there are also many top funding institutions in East Asia, including three from Japan and one each from Korea and China. This suggests that East Asia will have a solid competitive edge in the field of AI application in BCa in the future.

The bibliometric analysis of journals is essential. Through this section, researchers can learn about the journals gathering high level research about AI application in BCa and get access to research hotspots. In addition, this will be a guidance for researchers to submit their manuscripts. Table 3 summarizes the top 10 most published journals. These journals focus on the field of urology as well as oncology. EUROPEAN UROLOGY, BJU INTERNATIONAL and JOURNAL OF UROLOGY are the most critical journals in genitourinary systems as Q1 journals in JCR division. It is clear from Figure 5B that four journals, including EUROPEAN UROLOGY, BJU INTERNATIONAL, JOURNAL OF UROLOGY, and UROLOGY, are at the center of the citation relationship. Among them, EUROPEAN UROLOGY is, deservedly, an excellent journal for researchers in the field of AI application in BCa to study, which has the highest number of publications in this field with 11,848 citations. Although BJU INTERNATIONAL, JOURNAL OF UROLOGY, and UROLOGY did not achieve exceptionally high impact factors in 2021, they all had an average citation rate per paper of more than 28. Therefore, these journals have a strong influence within this field and are equally worthy of studying and finding hotspots for researchers.

Table 4 summarizes the top 10 authors in terms of publication volume. Seven of these researchers are from the United States. Wiklunk from Karolinska Institutet has the highest publication volume and total number of citations in the field of AI application in BCa. Guru, from Roswell Park Comprehensive Cancer Center, has the highest H-index in this field. Although Menon, Pruthi, and Peabody do not have more than 35 publications, they have an average of more than 82 citations per paper and are considered high-impact authors in the field. The four researchers focused their research on the following areas: robotic-assisted radical cystectomy, intracorporeal urinary diversion, quality of life, assisted chemotherapy and predictive modeling. A large portion of these are clinical studies related to robotic-assisted surgery. In addition, Guru has done more research on predictive modeling, urinary diversion and lymphadenectomy and has achieved more significant results (43–47). In recent years he has done much research on Neoadjuvant Chemotherapy (48–50). Menon and Peabody are colleagues at Henry Ford Hospital, and their research focuses on robotic surgery and urinary diversion surgery. This demonstrates the extremely high level of robotic surgery and research experience of Henry Ford Hospital. The three authors with the highest centrality are Guru, Rha, and Mottrie. Rha is from South Korea and has extensive collaborations with high-level researchers in Europe and the United States. For researchers in the field of AI application in BCa, we believe that learning from the work of these high-level researchers and actively participating in collaborative research projects will tremendously affect research capabilities.

Reference analysis is significant in finding highly cited publications and new research hotspots. Table 5 summarizes the top 10 most highly cited papers related to the field of AI application in BCa. They are mainly from three journals: EUROPEAN UROLOGY, BJU INTERNATIONAL and CLINICAL CANCER RESEARCH. The earliest published highly cited literature was Menon’s 2003 clinical study describing a technique of nerve-sparing robot-assisted radical cystoprostatectomy for patients with BCa (33). In the same year, Dyrskjot published a paper on BCa classification based on supervised algorithms in NATRUE GENETICS (35). These two papers are considered pioneering works in the field of AI application in BCa and have induced two different routes in the area at a later stage which are robotic surgery research and diagnostic and predictive models. Wu (2017), Parekh (2018), and Kamoun (2020) are highly cited documents that have emerged in recent years and are more representative of current research hotspots (51–53). These three papers mainly relate to radiomic prediction, robotic surgery evaluation, and consensus molecular classification. It is worth noting that these research directions are, in fact, a continuation of the two 2003 papers mentioned above. This suggests that research about the robotic surgery and AI diagnostic and predictive models has continued to emerge with new technologies over the past 23 years and is likely to grow explosively in the coming years.

Figure 7C summarizes the top 25 documents with the strongest citation bursts. To get more helpful research guidance, we focused more on the literature with a citation burst in 2022. Through burst detection and our screening, we chose to analyze the literature published by Robertson in 2017 and Parekh in 2018 (52, 54). Similar to previous findings, these two papers focus on robotic surgery research and AI diagnostic and predictive models. The study by Robertson et al., published in Cell, identified subgroups with differential epithelial-mesenchymal transition status, in situ carcinoma score, histological features, and survival by clustering the expression of mRNA, lncRNA and miRNA (54). And this was achieved based on many AI algorithms, such as the MutSig algorithm, ABSOLUTE algorithm, HLA typing algorithm, PathSeq algorithm, RBN algorithm, etc. And the study of Parekh et al., published in LANCET, evaluated robot-assisted radical cystectomy versus open radical cystectomy by the RAZOR trial (52). These articles still have citations explosion in 2022, indicating that robotic-assisted surgery and AI diagnostic and predictive models for BCa are still hot research topics. To find more research hotspots, we plotted a timeline of the references for the field of AI application in BCa ( Figure 7B ). Since 2013, the intracorporeal urinary diversion and open radical cystectomy have been the most explosive research hotspots on the surgery. Open radical cystectomy is a traditional surgical procedure for treatment of the BCa, but in recent years research on this procedure has still received frequent citations. From reading the relevant literature, we found that this burst is mainly due to the clinical comparison of open cystectomy versus robot-assisted cystectomy in recent years (52, 55, 56). And the robotic-assisted surgery has shown a significantly better postoperative quality of life and less blood loss (55). It can also be seen that artificial intelligence and deep learning-related documents have been widely emerging and cited since 2016. This was related to the rise of AI and the rapid advancement of related algorithms. AI diagnostic and predictive models need algorithmic improvements, which can explain the explosion of research in AI diagnostic models during this period.

We likewise performed a citation burst analysis of keywords ( Figure 8A ). Robotic surgical procedure, deep learning, convolutional neural network and artificial intelligence have had citation bursts in recent years. These are all keywords related to AI programming, which suggests that algorithms are of greater importance for future research about AI application in BCa. In addition, we performed a co-occurrence analysis of the keywords ( Figure 8B ). These keywords provide researchers with direct indication of research hotspots. Future hotspots in the field of AI application in BCa may be toward diagnostic and predictive models for BCa and improving robotic surgical procedures which require more programmers to work harder to achieve powerful algorithmic and higher accuracy. AI diagnostic and predictive models for BCa have been extensively studied in the last few years. Meanwhile, since 2019, convolutional neural networks and deep learning have triggered an intense citation explosion. This confirms that the AI model for pathology diagnosis and prediction are a hot research topic that has emerged in recent years. From 2019 to 2021, a large number of AI algorithm-based models for BCa diagnosis and prediction were developed, which include almost all aspects of BCa diagnosis and prediction: cytological diagnosis, histological diagnosis, prognosis prediction, automated grading, etc. (11, 57–60). Notably, testing of these models showed extremely high AUC and even detected low-grade BCa that pathologists missed. These provide essential guidance for radiation therapy and chemotherapy in managing the BCa. However, these models are not perfect, which means that they deserve to be further improved. According to the literature, the algorithms used in these models are mainly machine learning, deep learning and convolutional neural networks, and machine learning is currently one of the most mentioned techniques in the field of AI (61). Therefore, we recommend that researchers pay more attention to these techniques and improve the old or develop new models based on them. Regarding the diagnosis of BCa, a large number of studies have focused on the use of AI for pathological image analysis, but recently research on AI application in the cystoscopy has begun to emerge (62–64). One of the difficulties of this kind of research is to achieve the real-time detection of images (65), which may become a new research direction. Regarding the prognosis prediction of BCa, research is mainly focused on the prediction of survival, recurrence, treatment response, etc. (17) Among them, the application of deep learning is worth paying attention to, and the use of multimodal learning may become a future research direction (66). At the same time, the robotic surgery is still a high research hotspot today. The robotic surgery has always been a key application area of AI in medicine and thus has attracted the attention of many researchers (12–16, 67, 68). Notably, current research hotspots are turning to robotic-assisted radical cystectomy versus open radical cystectomy and intracorporeal urinary diversion (44, 45, 52, 55). These studies mainly focused on surgical outcomes and prognosis with little direct discussion about AI, but their results suggest that the robotic surgery is still worth improving, and focusing on the application of AI can provide ideas for this. An article focused on the hotspots and trends in the field of AI revealed that robots were a highly concerned topic in this field, including human-robot interaction, robot manipulation, robot grasp design, etc. (61) This suggests that AI has a broad application prospect in surgical robots, and researchers can make new explorations based on the above directions. Published research on AI and surgical robots is also worth referencing, including directions such as autonomous surgical robots, motion analysis, and presurgical planning (15, 67).

In short, the analysis of keywords brings us the following insights: the robot-assisted surgery for BCa radical treatment and urinary diversion is still a focus of the field of AI application in BCa, and the diagnostic and predictive models of AI for BCa can lead the subsequent explosion in this field. It is worth emphasizing that we also need to pay attention to the limitations of research on AI application in BCa, which can help us to be clearer about the future direction of improvement and development. Firstly, some studies of AI diagnostic and predictive models mentioned the need for prospective studies to further verify the reliability of these models (11, 57), suggesting that they cannot be used in current clinical practice. There are some other factors that hinder the integration of AI into clinical settings, such as the lack of generalizability across different datasets (58), the lack of availability and quality of datasets (65), and so on. In addition, information security issues may arise with the rapid development of AI and big data technology, which will further raise ethical issues and affect patient acceptance.

AI application in BCa is currently a trendy research area and research on it has been increasing since 2000. The hot research topics in this area include two main parts: AI models for the diagnosis and prediction of BCa and novel robotic-assisted surgery for BCa radicalization and urinary diversion. We venture to speculate that AI models for the diagnosis and prediction of BCa will be the next great hotspot in urology. In addition, improving and applying algorithms will be a strong driving force for development of this field. To achieve these, significant funding and the combined efforts of programmers and medical practitioners will be urgently needed.

Publicly available datasets were analyzed in this study. This data can be found here: https://webofscience.clarivate.cn/wos/

HJ and YJZ determined the topic and designed the study. WX and YZ searched and referred documents. HL, ZL and JL collected data. YJZ, WX, YZ, and TW analyzed data and mapped by the softwares. YJZ, WX and YZ drafted the manuscript. HJ was responsible for revision and correspondence.