Diagnosis of urinary tract infection based on artificial intelligence methods
Introduction
Urinary tract infection (UTI) is a common disease (approximately 95% of the total urinary tract infections) generally caused by E. coli bacteria reaching the bladder from the urinary tract and affecting at least half of the people throughout their lives [1], [2]. It may be observed as a simple inflammation of the bladder or a more severe inflammation of other urinary system organs. Inflammations involving only bladders that are not based on another underlying disease are called “simple infections” (cystitis) and all other urinary tract infections are called “complicated infections”. Cystitis is a bladder inflammation. This term is used as a pathological, bacteriological and cystoscopy term. Suddenly starting dysuria, pollacuria, urinary urgency and suprapubic pain are the syndromes of cystitis [3]. Cystitis is diagnosed with pollacuria, suprapubic pain, and burning sensation during urination, if the infection is detected only in the bladder. If the infection has retained the ureter, it is diagnosed as a nonspecific urethritis with diuresis, pollacuria, leukocyte urine findings. This diagnosis is also complicated since the symptoms are intertwined with one of the symptoms. The infection results in septicemia, death or at least the chronicity of the disease, if the organ in which the urinary tract infection is located is not treated properly. Under the top name of UTI, exist all kinds of inflammation on kidney, bladder, and urethra [1], [2].
150 million patients are reported to be diagnosed with UTI per year and $16 billion was spent on these patients [4]. In addition, the lifetime incidence of UTI counts 14,000 from 100,000 men [5]. In women, the three entities, the anus, the vagina mouth and the urethra, are very close to one another. In women, the three entities, the anus, the vagina mouth and the urethra, are very close to one another. Anal and vaginal area bacteria can easily become cystitis caused by passing from the urethra to the bladder. This rate is even higher due to the women urethral anatomy [6].
Routine urinalysis and urine cultures are utilized for the diagnosis of UTI. However, it is not easy to identify the source of the infection and the organ is infected. Depending on the characteristics of patients, it may be required to apply invasive methods such as urineletting separately from the bladder and kidney by placing a catheter in the ureters with the help of the cystoscopy. These invasive methods may cause more complications for the patient. For this reason, developing an AI model based on symptom and laboratory findings will protect the patient from possible complications.
UTI causes 20–30% of all systemic infections. In order to start treatment of the patient, a definite diagnosis is required whether cystitis or nonspecific urethritis is present. Since cystitis and nonspecific urethritis need to be treated with different drugs.
During the diagnosis, cystitis is suspected primarily because the main organ of the lower urinary tract is the bladder. The main symptoms of the acute cystitis are diarrhea, pollacuria, hematuria, leucocyturia, urinary urgency. Sometimes fever rarely participates in this history of acute urinary tract infections in children [7]. It is also accompanied by suprapubic pain. However, fever may also be observed in infections involving the bladder muscle (detrusor). Diabetes and recurrent UTI are also a risk factor for cystitis.
Patients with urethritis are applying to polyclinic with the symptoms of defluxion in urethra and dysuria. Among the urethritis agents, neisseria gonorrhoeae, chlamydia trachomatis, ureaplasma urealyticum and trichomonas vaginalis are the most common factors. However, except for gonorrhea infections, it is difficult to detect other factors in the laboratory environment [8]. In the clinic, it is both impractical and difficult to get results from the urethral drenage culture to diagnose the infections separately except for gonorrhea infections.
Artificial intelligence methods are successfully applied to situations like these where complexity exists. With the use of artificial intelligence in medicine, artificial intelligence programs can be created that can perform clinical diagnostic procedures and recommend treatment suggestions. There are a lot of successful applications of artificial intelligence in the field of medicine, such as classification and diagnosis of diseases, treatment recommendation, drug dosage determination, and so on [9], [10], [11], [12], [13], [14], [15], [16], [17], [18]. These applications use artificial intelligence methods such as SVM, ANN, DT, KNN and RF. These artificial intelligence methods are powerful tools that physicians can use to analyze, model, and understand complex clinical data in a variety of medical practice areas.
Furthermore, the use of AI methods in medical practice has become widely accepted. These methods include advantages such as (a) economically and non-linear modeling of large data sets and ease of optimization; (b) the predictive accuracy that has a potential to support clinical decision process; (c) providing an explanation using rule inference or sensitivity analysis which ease to disseminate the information [19].
With artificial intelligence methods, inputs that are independent variables of the system can form complex, nonlinear models by associating them with dependent predicted variable outputs. Artificial intelligence is the most important tool for modeling and decision making. Thus, in this study, various AI classification methods have been compared for diagnose supporting whether UTI with complicated symptoms is cystitis or nonspecific urethritis. Four different AI methods, DT, SVM, RF and ANN, which are frequently used in medical classification problems, have been used in this study [20], [21], [22], [23]. A comparative study of those will give us insight into the merits of the different AI methods when used on this problem.
Section snippets
Data set
In this study, the data set were used of the 59 patients 7–77 years old, 35 female and 24 male, who applied to the Necmettin Erbakan University Meram Medical Faculty Urology Polyclinic between the dates of 2016–2017. These patients were diagnosed and localized as urinary tract infection (cystitis) and nonspecific urethritis. All of these patients were diagnosed with the procedures such as anamnesis, clinical examination, urinalysis, and ultrasonography. A definitive diagnosis chart of the data
Preprocessing
The “normalization” of the data is important in terms of not corrupting the relationship between the variables, the accuracy of the analysis and the network performance. Normalization is a scaling process for each data in the dataset between the upper and lower bounds of the activation function used. The activation function used in the analysis is "sigmoid" function, during the implementation the data were normalized to the range of [0,1] by using the Eq. (9).
Furthermore,
Results
In this study, it is aimed to develop a model that predicts cystitis and nonspecific urethritis diseases with similar symptoms from urinary tract infections. For this purpose, anamnesis, full urine and ultrasonic examinations of 59 patients who applied to the Urology Clinic for the Meram Medical Faculty Hospital were collected and a UTI dataset was created. Four different artificial intelligence methods, namely DT, SVM, RF and ANN, which are frequently utilized in medical diagnosis systems were
Discussion
From a medical point of view, it is necessary to utilize a large number of symptom and laboratory findings in order to be able to diagnose UTI in the clinic. Even in some cases, subpubic ponction and urinary catheterization of both kidneys are used for localization. The two methods mentioned above are an invasive and complicated procedure. Especially, pollacuria and suprapubic symptoms are the most important indicators of cystitis. From this point of view, the results obtained from the DT
Conflict of interest
The authors of this manuscript declare that they have no financial and personal relationships with other people or organizations that could inappropriately influence their work.
Acknowledgments
The authors would like to thank Necmettin Erbakan University Meram Medical Faculty Urology Polyclinic for its contributions to the study. The authors also would like to thank the reviewers and the editors for their valuable comments and contributions that helped to increase the readability and organization of the present paper significantly.
References (54)
Urologic diseases in America project: trends in resource use for urinary tract infections in men
J. Urol.
(2005)Application of artificial neural networks to clinical medicine
Lancet
(1995)- et al.
Survey of utilisation of fuzzy technology in medicine and healthcare
Fuzzy Set Syst.
(2001) - et al.
Prognosis of prostate cancer by artificial neural networks
Expert Syst. Appl.
(2010) - et al.
Artificial neural networks in medical diagnosis
J. Appl. Biomed.
(2013) - et al.
The use of artificial neural networks in decision support in cancer: A systematic review
Neural Netw.
(2006) - et al.
Machine learning techniques for breast cancer computer aided diagnosis using different image modalities: a systematic review
Comput. Methods Prog. Biomed.
(2018) - et al.
Logistic regression and artificial neural network classification models: a methodology review
J. Biomed. Inform.
(2002) - et al.
Data mining method for listed companies’ financial distress prediction
Knowl. Based Syst.
(2008) - et al.
8-Classification: Basic Concepts
Simplifying decision trees
Int. J. man Mach. Stud.
Empirical characterization of random forest variable importance measures
Comput. Stat. Data Anal.
Automatic diagnosis of premature ventricular contraction based on Lyapunov exponents and LVQ neural network
Comput. Methods Progr. Biomed.
Artificial neural networks (the multilayer perceptron)—A review of applications in the atmospheric sciences
Atmos. Environ.
Management strategies for urinary and vaginal infections
Arch. Intern. Med.
Diagnosis of coliform infection in acutely dysuric women
N. Engl. J. Med.
Campbell's Urology
Urinary tract infections: disease panorama and challenges
J. Infect Dis.
Anatomy and physiology of the urinary tract: relation to host defense and microbial infection
Microbiol. Spectr.
Once-daily cefadroxil versus twice-daily cefaclor for treatment of acute urinary tract infections in children
J. Antimicrob. Chemother.
Chlamydia trachomatis infection and non-gonococcal urethritis in homosexual and heterosexual men in Edinburgh
Int. J. STD AIDS
Artificial intelligence in medicine
Ann. R. Coll. Surg. Engl.
Artificial neural networks in decision support systems
Handbook on Decision Support Systems 1
Using artificial neural network for the prediction of anemia seen in behcet disease
Energy Educ. Sci. Technol. Part A Energy Sci. Res.
Prediction of breast cancer using artificial neural networks
J. Med. Syst.
Introduction to medical applications of artificial intelligence
Medical Applications of Artificial Intelligence
Diagnosis of anemia in children via artificial neural network
Int. J. Intell. Syst. Appl. Eng.
Cited by (53)
Artificial Intelligence in Pediatric Urology
2024, Urologic Clinics of North AmericaDevelopment and validation of artificial intelligence models to predict urinary tract infections and secondary bloodstream infections in adult patients
2024, Journal of Infection and Public HealthA novel discrete learning-based intelligent methodology for breast cancer classification purposes
2023, Artificial Intelligence in MedicineDevelopment and validation of models for detection of postoperative infections using structured electronic health records data and machine learning
2023, Surgery (United States)Citation Excerpt :Previously published studies reported models for SSI detection with sensitivities between 0.65 and 0.79 and specificities between 0.92 and 1.00.9,11 A recent article showed that artificial intelligence detected UTIs in 59 patients using EHR data with 98% sensitivity and 100% specificity using artificial neural networks, but given this very small sample size, the use of a black box algorithm, and the lack of external validation, generalizability of these findings is questionable.43 The SSI model we presented is simple and is similar to that of Knepper et al, which suggests that it could also be used to augment NHSN SSI surveillance.31
The effect of machine learning algorithms in the prediction, and diagnosis of meningitis: A systematic review
2024, Health Science Reports