Technological progress in Artificial Intelligence (AI) has enabled clinicians to support decision-making and provide healthcare solutions by developing AI-based models that can extract data patterns. Medical data is extracted and fed into various AI models to obtain results that can aid in disease diagnosis and/or treatment.^[1,2] It is utilized as input for the development of therapies with diagnostic and prediction accuracies that are superior to conventional standards. Furthermore, virtual assistants^[3] have made it easier to converse with a technical system using Natural Language Processing (NLP). Responsible AI adoption necessitates that it be financially incentivized^[4] on a regular and sustainable basis.

Urinalysis comprises of a diverse array of assessments, encompassing chemical evaluations, microscopic investigations of urine, bacterial cultures, and molecular tests. The use of urinalysis that enjoys widespread recognition is its utility for determining pregnancy.^[5]Furthermore, urine is employed as a diagnostic tool for the identification of Non-alcoholic fatty liver disease (NAFLD), a condition that currently lacks non-invasive indicators for clinical diagnosis. The application of lipid molecular function of urinary Extracellular Vesicles shows potential in the detection of non-alcoholic steatohepatitis (NASH).^[6] The assessment of the DNA methylation status of genes and its potential association with the development of bladder cancer can also be conducted by analyzing urine sediment.^[7] The work is currently underway to assess the potential influence of improved urine screening tests in enhancing the sensitivity and efficacy of Hepatocellular carcinoma (HCC) detection. Additionally, the utilization of urine Circulating tumor DNA (ctDNA) as a screening test is also being explored.^[8] Urine has been identified as a viable non-invasive method for monitoring alterations in biochemical processes associated with the progression of cancer.^[9] The prospective diagnostic advancements offered by urine biopsy present notable benefits, which have the capacity to pave the path for novel approaches in therapy selection and the facilitation of precision disease therapies.^[10]

Artificial Intelligence has gained widespread acceptance in the field of urology for diagnosis and treatment, and with advancements in image processing, pattern recognition, and Machine Learning technology, conditions such as urinary tumors, urological calculi, and erectile dysfunction^[11] can now be diagnosed and treated with greater precision. While the majority of AI diagnosis and treatment is currently in the pre-clinical research phase, its present applications encompass automating cancer detection through the utilization of radiomic imaging and digitized tissue specimen images, aiding disease diagnosis by integrating patient clinical data, biomarkers, or gene expression, facilitating the planning of brachytherapy and radiation treatments, and employing robotic arms for automated surgeries.^[12] This review aims to investigate the potential research and application opportunities of AI in the field of urological diseases, specifically focusing on diagnosis and therapy.

1.1 AI combined with urine proteomics

Urine contains 15–150 mg of polypeptides and proteins. Changes in their composition can reflect physiological and pathological alterations. Using urine protein fluctuations, AI can detect diseases. Roux-Dalvai et al.^[13] used LC-MS and machine learning to create a peptide profile of urinary tract infection-causing bacteria and examine unknown urine samples using focused proteomics. Our approach detected infections faster and with 100% accuracy when applied to data over the clinical threshold of 1105 CFU/mL. (CFU: colony forming units). Lin et al.^[14] meta-analyzed 45 studies on urine albumin detection of diabetes and reported that DR had a sensitivity of 0.67, specificity of 0.78, and PDR of 0.99 in predicting DN. Model external verification accuracy was 0.875. Lucarelli et al.^[15] used neural networks to study genetic linkages to picture phenotypes utilizing renal histology and urinary proteomics. 315 handcrafted digital image features and 207 tubule characteristics through the HAIL pipeline and fully connected networks discovered differentially expressed proteins in urine that caused end-stage renal disease within 2 years of biopsy. For both glomeruli and tubules, RGB color values and PAS+ variation was more predictive of molecular profiles than other variables. This research demonstrates that AI can use urine proteome data to diagnose diseases non-invasively and accurately.

1.2 AI combined with urine metabolomics

Urine metabolomics data may be used to diagnose and cure diseases, however, the enormous number of compounds in urine presents challenges. Recently developed artificial intelligence technologies can evaluate and extract valuable information from this data. Urine metabolites and Machine Learning (ML) algorithms were used by Kouznetsova et al.^[16] to diagnose bladder cancer early and late. In distinguishing illness phases in the training set, their best model had an accuracy of 82.54% and an area under the precision-recall curve of 0.84. Gladding et al.^[17] used multi-omics and ML to assess data from heart failure patients with decreasing ejection fraction. Advanced Electrocardiogram (AECG) and Echo AI, done for over 5 minutes, showed a good connection with manually determined parameters including Left Ventricular End-Diastolic Volume, Left Ventricular End-Systolic Volume, and Left Ventricular Ejection Fraction. The study found an AUC of 0.95 and 95% CI: 0.85-0.99 for multi-omics and ML in heart failure evaluation. AI-ML algorithms and independent assessment of urinary metabolome data from cats with meloxicam-induced kidney injury were used by Broughton et al..^[18] They identified and validated a panel of metabolites that may aid clinicians in disease diagnosis and prognosis.

1.3 AI combined with urine RNomics

MicroRNAs regulate gene expression by affecting mRNA stability and translation. Urine contains tiny amounts of these short RNA molecules, making them excellent indicators for tumor formation. AI can detect aberrant microRNAs and diagnose forecast, and assess cancer treatment. Connell et al.^[19] used AI to predict prostate cancer using extracellular vesicle-extracted urine RNA. The model's 0.77 diagnostic accuracy in detecting intermediate to high-risk prostate cancer improved prognosis for active monitoring patients. Based on urine RNA levels, AI technology may be able to identify cancer and other disorders.

1.4 AI combined with urine cytopathology

Urine contains cells from both healthy and sick people, and most of them are important for diagnosing urinary system problems. White blood cells and pus cells in urine may indicate urinary tract infection. Red blood cell shape in urine helps identify glomerular disorders from other diseases. From the kidney to the urethra, epithelial cells line the urinary tract. Sanghvi et al.^[18] constructed a deep learning computational pipeline with many tiers of convolutional neural network models to process entire slide images and predict diagnoses. For high-grade urothelial cancer, the algorithm exhibited 79.5% sensitivity and 84.5% specificity. Yamasaki M et al.^[20] compared the Paris System (TPS) for reporting urine cytology, which focuses on high-grade urothelial melanoma, to the conventional systems (CS) and found no significant difference between the two. TPS and CS had 56.0 and 58.2 sensitivity, 97.8 and 91.2 specificity, and 93.3 and 87.9 positive predictive values, respectively. TPS's negative predictive value for HGUC was 80.0, significantly higher than CS's (66.4, P = 0.04). AI employing the VisioCyt test to improve bladder melanoma diagnosis using voided urine cytology was investigated by Lebret T et al.^[21] Two groups were studied: bladder melanoma diagnosis with varied histological grades and stages, and control cases with negative cystoscopy and cytology results. The VisioCyt test has a sensitivity of 84.9 compared to 43 for voided urine cytology. AI technology can improve urine cytopathology diagnostics, according to these studies.

1.5 AI combined with urinary function

Xiong et al.^[22] reported that Lower Urinary Tract Symptoms (LUTS) are common markers of urinary system problems, affecting both urine storage and micturition. The first include nocturia, urgency, and frequent urination, whereas the latter include dysuria, delayed urine inflow, and a thin urine stream. LUTS help diagnoses benign prostatic hyperplasia, urinary tract infection, neurogenic bladder, and others. Singh S et al.^[23] suggested assessing active and chronic lesions with a modified NIH effort and regularity score system. Due to inadequate data, 80 LN cases identified between June 2018 and April 2020 were retrospectively studied. Inactive urinary deposits, a 24-hour urine collection with PCR of 0.5 g/g, and normalized/stabilized renal function indicated complete remission (CR). The revised 2018 ISN/RPS classification and modified NIH score system described pathologic lesions. AI (low: 0-5; moderate: 6-11; high: 12-24) and CI scores categorized the instances (low: 0-2; moderate: 3-5; high: 6-12). Kaplan-Meier analysis determined event time. Multivariable Cox proportional hazard models analyzed CR prognostic factors. 50 instances (62.5%) attained CR after 8 months. High AI and moderate/high CI groups had decreased CR in the Kaplan-Meier analysis (p-value=0.001). Moderate and high CI scores and Glomerulosclerosis Scores were significant drivers of LN CR, with an HR of 0.088 (0.034-0.229) and p-value <0.001. Consequently, moderate and high CI scores were associated with a lower likelihood of CR in LN, with Glomerulosclerosis of CI being a major predictor. The results indicate that a urine function-based AI system can detect persistent lesions. Table 1 presents a comprehensive list of the articles that were meticulously reviewed in the preparation of this manuscript.

Table 1. Summary of recent studies from the literature on AI-driven urine analysis in clinical medicine.

Study	Objective	Dataset		Training features	Algorithms	Performance Measures				Outcome
Study	Objective	Training set	Testing set	Training features	Algorithms	Accuracy	Sensitivity	Specificity	AUC	Outcome
Hu, C. et al., 2021 ^[24]	An artificial Intelligence based interpretive reporting system for urine test results.	2 899 917 patients and 710 971 urine test data	199 abnormal urine test results	NA	AdaBoost algorithm	88.3%	80.0%	NA		> An intelligent result interpretive reporting system was established. > Distinguish the degree of abnormality in reports, had high accuracy, and provided personalized clinical decision-making data.
Yen-Chuan et al., 2022 ^[25]	Identify urothelial cancer candidate cells using whole-slide images (WSIs)	131 urine cytology slides	Performed on the same samples	NA	Deep- learning-based algorithm	NA	92.3%	98.9%	0.944	> The AI Algorithm assisted TPS-based reporting. > Provided AI-inferred WSIs and quantitative data
Ross J. Burton et al., 2019 ^[26]	To enable diagnostic services to concentrate on those in which there are true microbial infections	Urine samples with specimen (n = 225,207) Training (70%, n = 157,645)	30%, n = 67,562	NA	Combined models	65·65	95·2	60·93	0.749	> Shows the potential application of supervised ML models in improving efficiency when demand surpasses resources of public healthcare providers.
Adit B. Sanghvi et al., 2019 ^[27]	Development of an image algorithm that applies computational methods to digitized liquid-based urine cytology slides.	1436 cases that were of sufficient scan quality for analysis, 464 of which (32.3%) were diagnosed as SHGUC or HGUC.	790 cases	NA	A deep learning computational pipeline with multiple tiers of convolutional neural network models	84.2%	79.5%	84.5%	0.88	> Provides computer- assisted interpretation of urine cytology cases > ML technology – same as automated Papanicolaou test screening
Bendifallah et al., 2019 ^[28]	Predict the likelihood of endometriosis by using patient history, demographics, endometriosis phenotype, and treatment	1126 endometriosis patients, 608 controls	Validation process with a prospective cohort with n = 100	16 key clinical and patient-based symptom features	Logistic Regression, Random Forest, Decision Tree, eXtreme Gradient Boosting	NA	93%	92%	91%	> ML - replaced diagnostic laparoscopy > Patients - can contribute to shared decision making
van Bussel, M. J. P. et al., 2022 ^[29]	To analyze the determinants to accept a virtual assistant and use cases among cancer patients	8 former patients and 4 doctors	127 respondents		The unified theory of acceptance and use of technology (UTAUT)	Self- efficacy (ß = 0.792)	performance expectancy (ß = 0.399)	effort expectancy (ß = 0.258)	trust (ß = 0.210)	The support found for all UTAUT factors: > performance expectancy > effort expectancy >social influence > facilitating conditions
Zhu, Q. et al., 2022 ^[30]	Lipidomic identification of urinary extracellular vesicles for non-alcoholic steatohepatitis diagnosis	30 patients with NAFL and 28 patients with NASH	13 patients with NAFL and 12 patients with NASH	4 lipid biomarkers: FFA, LPC, FFA, PI	Random Forest Ensembling and decision trees	0.91	NA	NA	92%	> Effective distinction of NASH from NAFL
Venter J. M. E. et al., 2019 ^[31]	Comparison of an in-house real-time duplex PCR assay with commercial HOLOGIC® APTIMA assays for the detection of Neisseria gonorrhoeae and Chlamydia trachomatis in urine and extra-genital specimens	200 men	NA	16S rRNA of C. trachomatis	e HOLOGIC® APTIMA	100%	85.7%	100%	NA	The in-house duplex real-time PCR assay showed acceptable performance characteristics in comparison with the APTIMA® assays for the detection of extra-genital N. gonorrhea and C. trachomatis
Wirth M et al., 2018 ^[32]	A prospective observational pilot study to test the feasibility of a smartphone- enabled uChek© urinalysis device to detect biomarkers in urine indicative of preeclampsia / eclampsia	NA	350 enrolled pregnant women	Protein to creatinine ratio, microalbuminuria	decision support algorithms	70%	NA	NA	NA	> new smartphone-enabled medical device can be introduced with minimal training or additional resources
Schilling K. et al., 2020 ^[33]	Urine metallomics signature as an indicator of pancreatic cancer	Healthy (n = 46) and PDAC (n = 21) urine specimens were collected	NA	Zn isotopic composition, the potential of imbalance in trace elements	Mann–Whitney test	NA	95.2%	97.8%	0.995	> No significant variations for K, Li, Al, Rb, Ni, Cr, As, Mo, and Pb > The element concentrations of Na, Mg, Ca Fe, Cd, Cu, and Zn in urine differed between PDAC and healthy controls.
Kim, A. K. et al., 2022 ^[34]	Urine DNA biomarkers for hepatocellular carcinoma screening	609 patients from five medical centers	186 patients	Mutated TP53, and methylated RASSF1a, and GSTP1	A two-stage model was developed to combine AFP and urine panel as a screening test.	NA	79.6%	90%	NA	> Urine ctDNA has promising diagnostic utility in patients in HCC, especially in those with low AFP > Can be used as a potential non-invasive HCC screening test.
Parakh et al., 2019 ^[35]	Urinary stone detection on unenhanced CT images	535 patients (279 stones present; 256	100 scans (test data)	NA	Convolutional neural network (CNN)	>90%	NA	NA	NA	NA
Wang et al. ^[36]	To predict bladder cancer prognosis in terms of five-year overall and cancer-specific mortality using urinalysis	117 bladder cancer patients	NA	NA	Output-based transfer learning approach with least square support vector machine (LS-SVM)	5 years overall mortality Proposed classifier (v1): 76.97% Proposed classifier (v2): 76.18% 5-year cancer specific mortality Proposed classifier (v1): 74.85% Proposed classifier (v2)	5 years overall mortality Proposed classifier (v1): 78.48% Proposed classifier (v2): 78.29% 5-year cancer specific mortality Proposed classifier (v1): 90.26% Proposed classifier (v2): 92.38%	5 years overall mortality Proposed classifier (v1): 75.79% Proposed classifier (v2): 74.33% 5-year cancer-specific mortality Proposed classifier (v1): 38% Proposed classifier (v2): 31%	NA	NA
Gavriel et al.,2021 ^[37]	To predict a five-year prognosis of bladder cancer	78 patients diagnosed with MIBC	NA	NA	ML-based ensemble model	94.8%	89.5%	97.4%	NA	NA
Roux-Dalvai et al., 2019 ^[38]	Identified bacterial species causing UTIs quickly	190 samples including inoculated and non-inoculated urine	NA	82 peptides	RF and so on	100	NA	NA	0.98	NA
Kouznetsova et al., 2019 ^[39]	Identified early and late BCa	Metabolites obtained from publication (McDunn et al. 2015)	205 metabolites of early-stage BCa; 42 metabolites of late-stage BCa	All metabolites from the sources	ANN; SGD	72.00 (for early BCa); 65.45 (for late BCa)	NA	NA	NA	NA
Sapre et al., 2016 ^[40]	Constructed a BCa prediction model	30 patients with active cancer (recurrence); 30 non-recurred 21 benign controls	NA	6 parameters from the urine of patients	ANN	NA	NA	NA	0.961	NA
Connell et al., 2019 ^[41]	Found micro RNAs related to BCa and constructed a prostate cancer prediction mode	358 prostate patients	177 prostate patients	Urine-derived EV-RNA profiles	LASSO	NA	NA	NA	0.770	NA

2. Discussion

The integration of Artificial Intelligence within the medical sector holds great promise. In the forthcoming years, it is likely to completely revolutionize the medical world, whether through predicting outcomes or automating processes. Presently, diagnosis of diseases and analysis of medical images is performed manually, relying on the expertise and discretion of the diagnostician, resulting in a higher incidence of misdiagnosis, and missed cases in underprivileged areas or the hands of inexperienced physicians (Fig. 1). Table 2 delineates a comparative analysis between AI-based urine analysis and conventional methods, providing an insightful overview of the contrasting approaches. By introducing AI, it is hoped that this disparity will be alleviated, and misdiagnosis reduced. Moreover, with the growth of large medical datasets, AI can enhance the collection and processing of data, thereby increasing efficiency.

Fig. 1 Process flowchart of AI combined urine analysis.

Table 2. Comparison between AI based Urine analysis and traditional methods

Aspect	AI-based Urine Analysis	Traditional Diagnostic Methods
Methodology	Utilizes artificial intelligence algorithms and machine learning models for analysis	Manual examination and chemical reagents
Accuracy and Consistency	Highly accurate and consistent results	May vary based on technician expertise and subjectivity
Efficiency	Faster results and timely diagnosis	Time consuming
Data integration	Process and integrate vast amounts of data quickly, considering multiple parameters	Analysis based on a limited set of parameters
Resource requirements	Requires substantial computation resources and expertise	Less resource intensive
Dependency on training data	Accuracy depending on quality and diversity of training dataset	Results based in standardized reagents and established procedures
Bias	Potential for reflecting biases present in the training data	Less likely to reflect biases but subject to technician bias
Clinical Validation studies	Ongoing and evolving	Long history of use and validation studies
Practicality utility and advancements	Represents a significant advancement in diagnostic capabilities, promising faster, more precise diagnosis	Established and trusted in medical practice, continuous advancements in techniques and technologies

Urine data comprises a vast number of data points, that can be easily collected and is rich in information. By effectively utilizing urine data, it is possible to gain insight into the underlying pathogenic mechanisms of various diseases and aid doctors in their diagnoses and treatments. It is worth mentioning that urine collection is non-invasive and simple, and provides new diagnostic possibilities and ideas for diseases that currently require invasive testing methods.

Medical data is used to develop most AI products, which will improve as more data is collected and processed. Medical data comprises name, gender, age, past medical history, present ailment, and family history. Leaked data from this server or cloud potentially violate patient privacy. This makes it difficult to use AI in medicine while protecting patient privacy. More patient data, especially private data, is needed to improve AI solutions. Limiting health data collection restricts AI solutions while protecting privacy. AI research in medicine must find a balance between the two, but disease kind, country legislation, and societal attitudes complicate the matter.

Standardized protocols for collecting and processing urine samples play a crucial role in AI-based urine analysis, ensuring data consistency, reliability, and accuracy. Consistency and uniformity in collecting and processing urine samples are vital to generate reliable and comparable data across different patients, laboratories, or healthcare settings. The quality and integrity of urine samples are ensured through standardized procedures, enabling AI models to identify relevant patterns and produce accurate analyses. Moreover, standardized protocols facilitate collaboration, data sharing, and interoperability among different research groups and AI platforms. They provide a structured foundation for AI model development, leading to more effective algorithms that can be applied across various datasets. Additionally, standardized protocols allow for fair evaluations and benchmarking of AI models, aiding in the selection of the most effective solutions. Adhering to these standardized protocols also helps in complying with ethical guidelines and regulatory requirements, ensuring that urine sample collection and processing are conducted ethically and responsibly.

Several organizations and institutions have established guidelines and standards for urine sample collection and processing. The Clinical and Laboratory Standards Institute (CLSI) offers recommendations for urine sample collection, handling, and processing. They provide protocols for determining precision and accuracy in urine chemistry testing. The International Society for Environmental Epidemiology (ISEE) has published guidelines for urine sample collection and storage in epidemiological studies. These guidelines ensure proper procedures for collecting urine samples in research studies. Additionally, the National Institute for Occupational Safety and Health (NIOSH) offers recommendations for biological monitoring and the collection and analysis of urine samples in the field of occupational health. Adhering to these established standards and guidelines ensures that urine samples are collected and processed consistently, meeting quality and ethical requirements for AI-based analysis.

3. Datasets

The availability and accessibility of datasets in AI analysis for urine are critical for robust model development. However, there are substantial challenges related to data acquisition, quality, and quantity. Access to diverse and comprehensive datasets is fundamental, but publicly accessible, well-annotated urine datasets are often limited due to privacy concerns and ethical considerations, especially in healthcare. Acquiring consent and anonymizing data while adhering to ethical guidelines can be challenging. Moreover, data heterogeneity, varying formats, units, and parameters measured in urine analysis, adds complexity in data integration. Combining urine data with other health records or medical imaging data can enhance AI models, but differences in formats, standards, and privacy concerns make integration difficult. In terms of data quality, accurate and consistent labeling, handling noise, outliers, and missing data are significant challenges. Generating precise labels for urine analysis data requires expert domain knowledge, and inconsistent labeling can greatly affect model performance. Additionally, obtaining a sufficient quantity of labeled urine analysis data for training robust AI models is challenging, especially when certain conditions or diseases are rare, resulting in imbalanced datasets. Addressing these challenges necessitates collaborative efforts, adherence to privacy regulations, innovative approaches for data standardization, and the use of techniques like synthetic data generation to mitigate limitations and enhance dataset diversity and representativeness for AI analysis in urine diagnostics.

4. Data Privacy, Security, and Ethical Considerations

The integration of AI in medical research, especially in the context of patient urine data, raises significant concerns regarding data privacy, security, and ethical principles. To uphold patient confidentiality and adhere to ethical guidelines and data protection regulations, certain fundamental steps must be taken. Anonymization and encryption techniques should be applied to strip any personally identifiable information from patient data, ensuring privacy. Furthermore, employing secure storage systems with restricted access and access controls is essential to safeguard the data. Compliance with ethical guidelines involves obtaining informed consent from patients, outlining the purpose, risks, and benefits of the study. Ethical review board approval is crucial to ensure adherence to ethical standards, and compliance with regional data protection regulations is mandatory. Transparency, regular audits, training, and collaboration with legal experts are vital steps to guarantee data privacy, security, and ethical conduct throughout the research process.^[42-45]

5. Regulatory frameworks

The application of AI in healthcare, including urine analysis, is guided by several regulatory frameworks and ethical guidelines to ensure responsible and ethical use of AI technologies. One significant regulatory framework is the Health Insurance Portability and Accountability Act (HIPAA) in the United States, which sets the standards for safeguarding sensitive patient data, including information obtained through urine analysis. Compliance with HIPAA regulations is essential for AI applications in urine analysis to uphold privacy and security of patient data. Similarly, the General Data Protection Regulation (GDPR) in the European Union plays a crucial role, ensuring the protection and privacy of personal data, including health-related data from urine analysis. Any AI applications in urine analysis involving individuals within the EU must adhere to GDPR, encompassing aspects like consent, data anonymization, and secure data handling.

Moreover, the US Food and Drug Administration (FDA) provides guidelines and regulations for the development and deployment of AI-based medical devices, including those used in urine analysis. AI-powered urine analysis tools may require FDA approval to ensure compliance with safety and efficacy standards. The International Medical Device Regulators Forum (IMDRF) offers global guidance on software as a medical device (SaMD), which is relevant to AI-powered urine analysis tools, ensuring alignment with global regulatory requirements.

In addition to these regulatory frameworks, various medical associations, such as the American Medical Association (AMA) and the World Medical Association (WMA), provide ethical guidelines for the use of AI in medicine. These guidelines emphasize transparency, accountability, and the welfare of patients, all of which are directly applicable to AI-based urine analysis. Furthermore, the European Commission has established ethical guidelines for trustworthy AI, promoting principles of transparency, accountability, fairness, and human oversight. These guidelines are vital for guiding the development and deployment of AI technologies in healthcare, including the domain of AI in urine analysis.^[46-47]

Adherence to these regulatory frameworks and ethical guidelines is critical to ensuring that AI applications in urine analysis align with established standards related to data privacy, security, transparency, accountability, and patient rights. This compliance is essential for maintaining patient trust and well-being while leveraging the potential benefits of AI in healthcare.

6. Conclusion

In conclusion, AI is essential in maximising the benefits that may be gained from growing medical data. The processing and interpretation of urine analysis data, such as proteins, metabolites, and RNA, is the primary focus of our research, and we are applying artificial intelligence to this problem; this not only makes the diagnosis of diseases more accurate and quicker, but it also opens up new prospects for less intrusive and more straightforward disease detection. Although it is still in the early stages of development, AI-powered urine detection has already proven itself as a useful adjunct instrument for diagnosing conditions affecting the urinary system. However, there is still potential for improvement in the accuracy and specificity of diagnoses based on AI. In conjunction with urine testing, AI is poised to emerge as a significant mode of disease diagnosis and treatment and has become increasingly utilised for early detection, treatment, and follow-up monitoring of a wide range of illnesses. Table 3 outlines the clinical applications of AI in urine analysis, offering a comprehensive overview of how artificial intelligence is utilized in various medical contexts. Despite this, with the rapid progression of computer technology and medicine, AI is poised to emerge as a significant disease diagnosis and treatment mode.

Table 3. Clinical applications of ai in urine analysis.

Application	Description
AI combined with urine proteomics	Utilizes urine protein fluctuations to detect illness
AI combined with urine metabolomics	Uses urine metabolites and ML algorithms for diagnosis
AI combined with urine Rnomics	Detects aberrant microRNAs to diagnose and assess cancer
AI combined with urine cytopathology	Enhances urine cytopathology diagnostics with AI
AI combined with urinary function	Detects urinary system problems using AI

7. Future directions

In the rapidly evolving field of AI-powered urine analysis, several exciting future directions and emerging trends are reshaping its role in healthcare. Beyond the diseases previously discussed, AI holds promise for early detection of various cancers, including bladder, prostate, and kidney cancers, by analyzing specific biomarkers and cellular changes in urine samples. Additionally, it can be instrumental in monitoring chronic conditions like diabetes and hypertension through continuous analysis of relevant urine markers, guiding better disease management and tailored treatment plans. AI algorithms are also poised to assess drug efficacy, predict metabolic and systemic disorders, and provide insights into personalized nutrition and wellness based on urine analysis. Furthermore, AI's potential in remote patient monitoring, predicting antimicrobial resistance in urinary tract infections, and integrating with electronic health records offers a glimpse into a future where AI-powered urine analysis plays a critical role in diagnostics, drug development, and optimizing healthcare outcomes.

Conflict of Interest

There is no conflict of interest.

Supporting Information

Not applicable.

References

[1] L. Petersson, I. Larsson, J. M. Nygren, P. Nilsen, M. Neher, J. E. Reed, D. Tyskbo, P. Svedberg, Challenges to implementing artificial intelligence in healthcare: a qualitative interview study with healthcare leaders in Sweden, BMC Health Services Research, 2022, 22, 1-16, doi: 10.1186/s12913-022-08215-8.

[2] B. Sivajohan, M. Elgendi, C. Menon, C. Allaire, P. Yong, M. A. Bedaiwy, Clinical use of artificial intelligence in endometriosis: a scoping review, Npj Digital Medicine, 2022, 5, 109, doi: 10.1038/s41746-022-00638-1.

[3] M. J. P. Bussel, G. J. Odekerken-Schröder, C. Ou, R. R. Swart, M. J. G. Jacobs, Analyzing the determinants to accept a virtual assistant and use cases among cancer patients: a mixed methods study, BMC Health Services Research, 2022, 22, 1-23, doi: 10.1186/s12913-022-08189-7.

[4] M. D. Abràmoff, C. Roehrenbeck, S. Trujillo, J. Goldstein, A. S. Graves, M. X. Repka, E. Silva, A reimbursement framework for artificial intelligence in healthcare, Npj Digital Medicine, 2022, 5, 72, doi: 10.1038/s41746-022-00621-w.

[5] N. O. Eltai, H. Alhussain, S. Doiphode, A. Al Thani, H. Yassine, Urine tests for diagnosis of infectious diseases and antibiotic-resistant pathogens. Pathogenic Bacteria: IntechOpen, 2020, doi: 10.5772/intechopen.89231.

[6] Q. Zhu, H. Li, Z. Ao, H. Xu, J. luo, C. Kaurich, R. Yang, P.-W. Zhu, S.-D. Chen, X.-D. Wang, L.-J. Tang, G. Li, H. Ou-Yang, M.-H. Zheng, H.-P. Li, F. Liu, Lipidomic identification of urinary extracellular vesicles for non-alcoholic steatohepatitis diagnosis, Journal of Nanobiotechnology, 2022, 20, 1-11, doi: 10.1186/s12951-022-01540-4.

[7] M. El azzouzi, H. El ahanidi, C. Hafidi Alaoui, I. Chaoui, L. Benbacer, M. Tetou, I. Hassan, M. Bensaid, M. Oukabli, A. Ameur, A. Al bouzidi, M. El mzibri, M. Attaleb, Exploring urine sediments as a non-invasive method for DNA methylation detection in bladder cancer, African Journal of Urology, 2022, 28, 1-5, doi: 10.1186/s12301-022-00298-3.

[8] A. K. Kim, J. P. Hamilton, S. Y. Lin, T.-T. Chang, H.-W. Hann, C.-T. Hu, Y. Lou, Y.-J. Lin, T. P. Gade, G. Park, H. Luu, T.-J. Lee, J. Wang, D. Chen, M. G. Goggins, S. Jain, W. Song, Y.-H. Su, Urine DNA biomarkers for hepatocellular carcinoma screening, British Journal of Cancer, 2022, 126, 1432-1438, doi: 10.1038/s41416-022-01706-9.

[9] K. Schilling, F. Larner, A. Saad, R. Roberts, H. M. Kocher, O. Blyuss, A. N. Halliday, T. Crnogorac-Jurcevic, Urine metallomics signature as an indicator of pancreatic cancer, Metallomics, 2020, 12, 752-757, doi: 10.1039/d0mt00061b.

[10] C. K. Chen, J. Liao, M. S. Li, B. L. Khoo, Urine biopsy technologies: cancer and beyond, Theranostics, 2020, 10, 7872-7888, doi: 10.7150/thno.44634.

[11] J. Chen, D. Remulla, J. H. Nguyen, A. Dua, Y. Liu, P. Dasgupta, A. J. Hung, Current status of artificial intelligence applications in urology and their potential to influence clinical practice, BJU International, 2019, 124, 567-577, doi: 10.1111/bju.14852.

[12] F. Qin, J. Yuan, Research status and trend of artificial intelligence in the diagnosis of urinary diseases, Sheng wu yi xue gong cheng xue za zhi = Journal of biomedical engineering = Shengwu yixue gongchengxue zazhi, 2020, 37, 230–235, doi: 10.7507/1001-5515.201910055.

[13] F. Roux-Dalvai, C. Gotti, M. Leclercq, M.-C. Hélie, M. Boissinot, T. N. Arrey, C. Dauly, F. Fournier, I. Kelly, J. Marcoux, J. Bestman-Smith, M. G. Bergeron, A. Droit, Fast and accurate bacterial species identification in urine specimens using LC-MS/MS mass spectrometry and machine learning, Molecular & Cellular Proteomics, 2019, 18, 2492-2505, doi: 10.1074/mcp.tir119.001559.

[14] S. Lin, Y. Guo, Value analysis of using urinary microalbumin in artificial intelligence medical institutions to detect early renal damage in diabetes, Journal of Healthcare Engineering, 2021, 2021, 1-13, doi: 10.1155/2021/6678454.

[15] N. Lucarelli, D. Yun, D. Han, B. Ginley, K. C. Moon, A. Rosenberg, J. E. Tomaszewski, S. S. Han, P. Sarder, Computational integration of renal histology and urinary proteomics using neural networks, Medical Imaging 2022: Digital and Computational Pathology. February 20-March 28, 2022. San Diego, USA. SPIE, 2022, doi: 10.1117/12.2613500.

[16] V. L. Kouznetsova, E. Kim, E. L. Romm, A. Zhu, I. F. Tsigelny, Recognition of early and late stages of bladder cancer using metabolites and machine learning, Metabolomics, 2019, 15, 1-15, doi: 10.1007/s11306-019-1555-9.

[17] P. A. Gladding, S. Loader, K. Smith, E. Zarate, S. Green, S. Villas-Boas, P. Shepherd, P. Kakadiya, W. Hewitt, E. Thorstensen, C. Keven, M. Coe, B. Nakisa, T. Vuong, M. N. Rastgoo, M. Jüllig, V. Starc, T. T. Schlegel, Multiomics, virtual reality and artificial intelligence in heart failure, Future Cardiology, 2021, 17, 1335-1347, doi: 10.2217/fca-2020-0225.

[18] L. E. Broughton-Neiswanger, S. M. Rivera-Velez, M. A. Suarez, J. E. Slovak, P. E. Piñeyro, J. K. Hwang, N. F. Villarino, Urinary chemical fingerprint left behind by repeated NSAID administration: discovery of putative biomarkers using artificial intelligence, PLoS One, 2020, 15, e0228989, doi: 10.1371/journal.pone.0228989.

[19] S. P. Connell, M. Yazbek-Hanna, F. McCarthy, R. Hurst, M. Webb, H. Curley, H. Walker, R. Mills, R. Y. Ball, M. G. Sanda, K. L. Pellegrini, D. Patil, A. S. Perry, J. Schalken, H. Pandha, H. Whitaker, N. Dennis, C. Stuttle, I. G. Mills, I. Guldvik, C. Parker, D. S. Brewer, C. S. Cooper, J. Clark, M. G. U. B. Consortium, A four-group urine risk classifier for predicting outcomes in patients with prostate cancer, BJU International, 2019, 124, 609-620, doi: 10.1111/bju.14811.

[20] M. Yamasaki, R. Taoka, K. Katakura, T. Matsunaga, N. Kani, T. Honda, S. Harada, Y. Tohi, Y. Matsuoka, T. Kato, H. Okazoe, H. Tsunemori, N. Ueda, R. Haba, M. Sugimoto, The Paris System for reporting urinary cytology improves the negative predictive value of high-grade urothelial carcinoma, BMC Urology, 2022, 22, 1-6, doi: 10.1186/s12894-022-01005-8.

[21] T. Lebret, G. Pignot, M. Colombel, L. Guy, X. Rebillard, L. Savareux, M. Roumigue, S. Nivet, M. Coutade Saidi, E. Piaton, C. Radulescu, Artificial intelligence to improve cytology performances in bladder carcinoma detection: results of the VisioCyt test, BJU International, 2022, 129, 356-363, doi: 10.1111/bju.15382.

[22] Y. Xiong, Y. Zhang, X. Li, F. Qin, J. Yuan, The prevalence and associated factors of lower urinary tract symptoms suggestive of benign prostatic hyperplasia in aging males, The Aging Male, 2020, 23, 1432-1439, doi: 10.1080/13685538.2020.1781806.

[23] S. Singh, H. C. Sreenidhi, Prognostic value of modified National Institute of Health activity and chronicity scoring in determining complete renal response in newly diagnosed lupus nephritis: a retrospective single centre study, International Urology and Nephrology, 2022, 54, 2075-2082, doi: 10.1007/s11255-022-03112-9.

[24] C. Hu, D. Yang, Z. Ye, Z. Liu, Y. Chen, Interpreting report of urinalysis based on intelligent data and machine learning, Chinese Journal of Laboratory Medicine, 2021, 44, 524-531, doi: 10.3760/cma.j.cn114452-20201016-00776

[25] Y.-C. Ou, T.-Y. Tsao, M.-C. Chang, Y.-S. Lin, W.-L. Yang, J.-F. Hang, C.-B. Li, C.-M. Lee, C.-H. Yeh, T.-J. Liu, Evaluation of an artificial intelligence algorithm for assisting the Paris System in reporting urinary cytology: a pilot study, Cancer Cytopathology, 2022, 130, 872-880, doi: 10.1002/cncy.22615.

[26] R. J. Burton, M. Albur, M. Eberl, S. M. Cuff, Using artificial intelligence to reduce diagnostic workload without compromising detection of urinary tract infections, BMC Medical Informatics and Decision Making, 2019, 19, 1-11, doi: 10.1186/s12911-019-0878-9.

[27] A. B. Sanghvi, E. Z. Allen, K. M. Callenberg, L. Pantanowitz, Performance of an artificial intelligence algorithm for reporting urine cytopathology, Cancer Cytopathology, 2019, 127, 658-666, doi: 10.1002/cncy.22176.

[28] J. S. Moore, P. R. Gibson, R. E. Perry, R. E. Burgell, Endometriosis in patients with irritable bowel syndrome: specific symptomatic and demographic profile, and response to the low FODMAP diet, Australian and New Zealand Journal of Obstetrics and Gynaecology, 2017, 57, 201-205, doi: 10.1111/ajo.12594.

[29] M. J. P. Bussel, G. J. Odekerken-Schröder, C. Ou, R. R. Swart, M. J. G. Jacobs, Analyzing the determinants to accept a virtual assistant and use cases among cancer patients: a mixed methods study, BMC Health Services Research, 2022, 22, 1-23, doi: 10.1186/s12913-022-08189-7.

[30] Q. Zhu, H. Li, Z. Ao, H. Xu, J. luo, C. Kaurich, R. Yang, P.-W. Zhu, S.-D. Chen, X.-D. Wang, L.-J. Tang, G. Li, H. Ou-Yang, M.-H. Zheng, H.-P. Li, F. Liu, Lipidomic identification of urinary extracellular vesicles for non-alcoholic steatohepatitis diagnosis, Journal of Nanobiotechnology, 2022, 20, 1-11, doi: 10.1186/s12951-022-01540-4.

[31] J. M. E. Venter, P. M. Mahlangu, E. E. Müller, D. A. Lewis, K. Rebe, H. Struthers, J. McIntyre, R. S. Kularatne, Comparison of an in-house real-time duplex PCR assay with commercial HOLOGIC® APTIMA assays for the detection of Neisseria gonorrhoeae and Chlamydia trachomatis in urine and extra-genital specimens, BMC Infectious Diseases, 2019, 19, 1-7, doi: 10.1186/s12879-018-3629-0.

[32] M. Wirth, N. Biswas, S. Ahmad, H. S. Nayak, A. Pugh, T. Gupta, I. Mahmood, A prospective observational pilot study to test the feasibility of a smartphone enabled uChek© urinalysis device to detect biomarkers in urine indicative of preeclampsia/eclampsia, Health and Technology, 2019, 9, 31-36, doi: 10.1007/s12553-018-0248-0.

[33] K. Schilling, F. Larner, A. Saad, R. Roberts, H. M. Kocher, O. Blyuss, A. N. Halliday, T. Crnogorac-Jurcevic, Urine metallomics signature as an indicator of pancreatic cancer, Metallomics, 2020, 12, 752-757, doi: 10.1039/d0mt00061b.

[34] A. K. Kim, J. P. Hamilton, S. Y. Lin, T.-T. Chang, H.-W. Hann, C.-T. Hu, Y. Lou, Y.-J. Lin, T. P. Gade, G. Park, H. Luu, T.-J. Lee, J. Wang, D. Chen, M. G. Goggins, S. Jain, W. Song, Y.-H. Su, Urine DNA biomarkers for hepatocellular carcinoma screening, British Journal of Cancer, 2022, 126, 1432-1438, doi: 10.1038/s41416-022-01706-9.

[35] A. Parakh, H. Lee, J. H. Lee, B. H. Eisner, D. V. Sahani, S. Do, Urinary stone detection on CT images using deep convolutional neural networks: evaluation of model performance and generalization, Radiology: Artificial Intelligence, 2019, 1, e180066, doi: 10.1148/ryai.2019180066.

[36] Y. S. Liu, R. Thaliffdeen, S. Han, C. Park, Use of machine learning to predict bladder cancer survival outcomes: a systematic literature review, Expert review of pharmacoeconomics & outcomes research, 2023, 23, 761-771, doi: 10.1080/14737167.2023.2224963

[37] C. G. Gavriel, N. Dimitriou, N. Brieu, I. P. Nearchou, O. Arandjelović, G. Schmidt, D. J. Harrison, P. D. Caie, Assessment of immunological features in muscle-invasive bladder cancer prognosis using ensemble learning, Cancers, 2021, 13, 1624, doi: 10.3390/cancers13071624.

[38] F. Roux-Dalvai, C. Gotti, M. Leclercq, M.-C. Hélie, M. Boissinot, T. N. Arrey, C. Dauly, F. Fournier, I. Kelly, J. Marcoux, J. Bestman-Smith, M. G. Bergeron, A. Droit, Fast and accurate bacterial species identification in urine specimens using LC-MS/MS mass spectrometry and machine learning, Molecular & Cellular Proteomics, 2019, 18, 2492-2505, doi: 10.1074/mcp.tir119.001559.

[39] V. L. Kouznetsova, E. Kim, E. L. Romm, A. Zhu, I. F. Tsigelny, Recognition of early and late stages of bladder cancer using metabolites and machine learning, Metabolomics, 2019, 15, 1-15, doi: 10.1007/s11306-019-1555-9.

[40] N. Sapre, G. MacIntyre, M. Clarkson, H. Naeem, M. Cmero, A. Kowalczyk, P. D. Anderson, A. J. Costello, N. M. Corcoran, C. M. Hovens, A urinary microRNA signature can predict the presence of bladder urothelial carcinoma in patients undergoing surveillance, British Journal of Cancer, 2016, 114, 454-462, doi: 10.1038/bjc.2015.472.

[41] S. P. Connell, M. Yazbek-Hanna, F. McCarthy, R. Hurst, M. Webb, H. Curley, H. Walker, R. Mills, R. Y. Ball, M. G. Sanda, K. L. Pellegrini, D. Patil, A. S. Perry, J. Schalken, H. Pandha, H. Whitaker, N. Dennis, C. Stuttle, I. G. Mills, I. Guldvik, C. Parker, D. S. Brewer, C. S. Cooper, J. Clark, M. G. U. B. Consortium, A four-group urine risk classifier for predicting outcomes in patients with prostate cancer, BJU International, 2019, 124, 609-620, doi: 10.1111/bju.14811.

[42] J. He, S. L. Baxter, J. Xu, J. Xu, X. Zhou, K. Zhang, The practical implementation of artificial intelligence technologies in medicine, Nature Medicine, 2019, 25, 30-36, doi: 10.1038/s41591-018-0307-0.

[43] Y. Q. Yu, H. F. Xu, J. Q. Hao, Y. L. Qiao, Zhongguo yi xue ke xue yuan xue bao. Acta Academiae Medicinae Sinicae, 2020, 42, 128-131. https://doi.org/10.3881/j.issn.1000-503X.10961.

[44] K. H. Keskinbora, Medical ethics considerations on artificial intelligence, Journal of Clinical Neuroscience, 2019, 64, 277-282, doi: 10.1016/j.jocn.2019.03.001.

[45] J. Tan, F. Qin, J. Yuan, Current applications of artificial intelligence combined with urine detection in disease diagnosis and treatment. Translational andrology and urology, 2021, 10, 1769-1779, doi: 10.21037/tau-20-1405.

[46] W. N. Price II, I. G. Cohen, Privacy in the age of medical big data, Nature Medicine, 2019, 25, 37-43, doi: 10.1038/s41591-018-0272-7.

[47] Véliz C. Medical privacy and big data: A further reason in favour of public universal healthcare coverage. Oxford (UK): Oxford University Press, 2019.