Artificial intelligence and varicocelectomy: A new horizon for patient management? A narrative review

Abstract

Varicocele is a common entity found in 15% of men and is the most common reversible cause of male factor infertility. Guidelines have been developed to guide urologists in deciding which patients would benefit from varicocelectomy. Yet studies published over the last decade showed the emergence of predictors of success of varicocelectomy using nomograms and other predictive models with statistical analysis. The emergence of artificial intelligence (AI) and machine learning revolutionized the clinician's approach to medicine. The virtual branch of AI, represented by machine learning, has been a very exciting topic for clinicians and researchers over the last years, especially after the launching of ChatGPT-3.5. Urology has been at the forefront of integrating advances in AI into its everyday practice. We aim to shed light on the present literature describing the use of AI in predicting the outcomes of varicocelectomy. Machine learning is being used to predict the improvement in semen parameters after varicocelectomy. These algorithms are derived from studies and data present in the literature and predictive models developed throughout the last two decades and have a superior performance to that of traditional nomograms. However, these models require further research and validation but are anticipated to surpass the accuracy of all current resources, setting forward a new era of varicocele workup and management in the years to come. This paper offers a wide review on the current evidence behind varicocele surgery and the integration of AI in medicine, urology and its use in predicting improvement in sperm parameters post-varicocelectomy.

Graphical abstract

Keywords

artificial intelligence / artificial neural network / machine learning / varicocele / varicocelectomy

1 INTRODUCTION

Varicocele is defined as an identifiable varicose vein that is a consequence of an abnormal dilatation of the spermatic vein and, subsequently, the pampiniform plexus within the spermatic cord[1]. Overall, it is identified in 15% of all healthy men[2]. It is found in up to 25% of men with abnormal semen parameters, but only 12% of men with normal semen parameters, which suggests its role in affecting male fertility and normal sperm production[3]. In fact, it is the most common reversible cause of infertility affecting around 35%–40% of men with primary infertility and 80% of men with secondary infertility[2,4].

The causes and the pathophysiology of varicocele remain on unsteady grounds. However, many theories have been proposed to elucidate its impact on male fertility. These include elevation of intra-testicular temperature[5], hypoxia[6], increase in oxidative stress[5], increased venous pressure[7], hormonal imbalance, in particular the decrease in testosterone[8], and the reflux of toxic metabolites to the testicle from adrenal or renal origin[9].

The mainstay for diagnosis of varicocele remains the physical examination, which is to be performed in a warm room, both in the supine and standing positions. The scrotum should first be visualized for the detection of a visible varicocele, then palpated. Finally, the examination should include an assessment of the veins with the patient performing the Valsalva maneuver[10]. Various grading systems have been proposed over the years to help standardize varicocele grade documentation, with the Dubin and Amelar grading system[11] most commonly utilized as follows: Grade 1, defined as the presence of varicocele expressed upon Valsalva maneuver; Grade 2, defined as the presence of varicocele upon palpation without the Valsalva maneuver, but not visible; and Grade 3, which represents varicocele that is visible through the scrotal skin while the patient is standing[10]. In addition to assessing for the presence or absence of a varicocele, another important aspect of the physical examination, especially in the context of fertility, is the assessment of testicular size. This can be done either using an orchidometer, or through the more reliable trans-scrotal ultrasonography, where testicular size and size discrepancy can be properly assessed, and where a radiological confirmation of varicocele can be obtained[12].

Varicocelectomy is a surgical procedure in which the internal spermatic veins are ligated and is a commonly performed procedure in infertile men. Several guidelines have been developed to guide urologists in the proper selection of patients with varicoceles who may benefit from varicocelectomy[13]. Furthermore, several authors have also developed nomograms and predictive models using logistic regression to help predict the outcomes of such a procedure. However, despite all this, it is still difficult for us to predict whom from our patients actually need surgery, and what the potential benefit may be.

The worldwide surge in the adoption of artificial intelligence (AI) and the race to incorporate its integration into all different fields confirms its increased importance as an essential and unseen component of everyday life[14]. The application of AI in medicine follows two different paths: physical and virtual[15]. Physical artificial intelligence started with the invention of robots back in 1921, followed soon after by increasingly sophisticated medical devices, metamorphosing in the early 2000s to surgical robotic systems[15]. These systems were first developed by the American Intuitive Surgical company, named Da Vinci, FDA-approved in the year 2000, and with now more than 7500 systems being used worldwide[15].

The virtual branch of AI, represented by machine learning, has been the hot topic for the past few years, marking international media recognition and an increased interest after the launching of the refined and more robust ChatGPT-3.5, a chatbot model introduced in November 2022 by OpenAI based on a large language model that was pretrained using large data allowing it to mimic human intelligence[15,16]. Machine learning uses mathematical algorithms and statistical models to learn from patterns in data, and therefore able to learn and reproduce these patterns when fed with a new set of data[15]. Machine learning includes the concept of deep learning, which goes a step further by learning from data through complex artificial neural networks that mimic the human brain[15].

Although marketing, customer support, data management and many other fields have seen an extreme reliance and application of machine learning models to increase efficiency; healthcare and medical science were relatively restricted in comparison, undermining a great potential of a revolutionized approach to patient care[16].

The integration of AI in medicine has also reached urology in all its various fields, with AI models being used in urologic oncology to help in the diagnosis of prostate cancer[16], and in robotic surgery where AI is used to facilitate intraoperative steps[17].

Nomograms are a simplified entity in which variables may be inserted in a certain model to obtain predictive values of a certain outcome through a two-dimensional diagram with certain variables on one side and the predicted outcome on the other[18]. Numerous nomograms and prediction models have been published in the literature such as those developed by Maimaitiming et al.[19], and Kandevani et al.[20]. These models stress the importance of predictive factors that would assist urologists in taking the decision of operating on varicoceles.

Interestingly, the literature now contains several articles in which AI may be used to help predict the outcomes of varicocelectomy and thereby potentially replace nomograms as tools of predictive outcomes.

The purpose of this article is to review contemporary literature on the use of machine learning and AI in predicting the outcomes of varicocelectomy.

2 METHODS

A thorough literature review was done on Medline, PubMed, and Embase using the following search criteria

On Medline: (exp Artificial Intelligence/OR ((deep* OR machin*) adj3 (learn* OR languag*)).mp. OR ChatGPT.mp. OR (artificial* adj3 intellig*).mp. OR (algorithm*.mp. OR AI.mp.) AND (exp Varicocele/OR varicocel*.mp. OR (pampiniform adj3 plexus).mp.)

On PubMed: ((“Artificial Intelligence”[Mesh] OR “Artificial Intelligence”[tw] OR (“deep learning”[tw] OR “machine learning”[tw]) OR ChatGPT[tw] OR AI[tw]) AND (“Varicocele”[Mesh] OR varicocel*[tw] OR “pampiniform plexus”[tw]))

On Embase: (‘artificial intelligence’/exp OR ‘artificial intelligence’ OR ‘deep learning’:ti,ab,kw OR ‘deep learning’/exp OR ‘machine learning’/exp OR ‘artificial neural network’/exp OR ‘machine learning’:ti,ab,kw OR chatgpt:ti,ab,kw OR ‘chatgpt’/exp OR ai:ti,ab,kw) AND (‘varicocele’/exp OR ‘varicocele’ OR varicocel*:ti,ab,kw OR ‘pampiniform plexus’:ti,ab,kw)

The titles and abstracts of all the papers were reviewed by two independent investigators, and only papers relevant to the latter topic were selected. Disagreements were resolved by consensus with the corresponding author. All selected articles were thoroughly reviewed and included in this study. No unpublished studies were described.

Articles included were studies that addressed machine learning and varicocelectomy. Articles addressing robotics and varicocelectomy were excluded. All retrieved articles were in English, and therefore, no language barriers were encountered. References of all papers were scanned for additional relevant articles, although none was found to meet the inclusion criteria as shown in Figure 1. The risk of bias was not assessed for the purpose of this narrative review.

3 RESULTS

Following an extensive literature review of Medline, PubMed, and Embase databases on the intersection of virtual AI and varicoceles, 182 articles were retrieved. Forty-one articles were triplicates among PubMed, Medline and Embase. 28 articles were duplicates between Medline and Embase. The rest of the 72 articles were screened, and 37 were excluded as they did not address varicoceles, the primary topic of this article. The remaining 35 articles were then assessed for eligibility, and 30 were excluded as they tackled robotics in relation to varicocele management, with no mention of virtual artificial intelligence and machine learning. The remaining five articles were then included in the review.

4 VARICOCELE SURGERY: BENEFIT, OUTCOMES AND CALCULATORS

Varicocele repair has been a hot topic of debate over the last several decades, especially as it pertains to the selection of the surgical candidates that are to undergo varicocelectomy. Nevertheless, guidelines have been developed to allow urologists to properly select patients that may benefit from varicocelectomy. According to the most updated European Association of Urology (EAU) guidelines, these patients are those that satisfy one of the following criteria (strong recommendation): adolescents with testicular size discrepancy of more than 2 mL or 20% confirmed on two subsequent visits 6 months apart risk ratio (RR). Patients having abnormal semen parameters and inability to conceive with the female partner having good ovarian reserve. There is still a weak recommendation to perform varicocelectomy for patients with raised DNA fragmentation index with unexplained infertility and having failed assisted reproductive techniques, including recurrent pregnancy loss, failure of embryogenesis and implantation[21].

Current evidence suggests that microsurgical subinguinal varicocelectomy is the most effective surgical technique in the treatment of varicoceles as it significantly improves pregnancy rates compared to other surgical techniques, as shown in a Cochrane review done by Persad et al, with RR intrauterine insemination (IUI) = 1.18[22,23]. It provides superiority in terms of improvement in sperm concentration and sperm motility, as well as decreased complication and recurrence rates as compared to open and laparoscopic varicocelectomy as shown in a meta-analysis done by Wang et al., and in a randomized controlled clinical trial by Bryniarski et al.[21,23,24]. However, this procedure requires specialized microsurgical training to perform effectively and achieve the best outcomes for patients, which may not be available to everyone.

A more recent meta-analysis by Argawal et al., showed significantly improved semen parameters compared to untreated controls post-operatively in terms of sperm concentration, total sperm count, progressive sperm motility, normal sperm morphology, but no significant difference for semen volume[25]. The authors concluded that varicocelectomy is effective in improving semen parameters in patients with clinically significant varicocele.

Varicocele repair has also shown benefit in other clinical scenarios such as patients with non-obstructive azoospermia (NOA). Multiple studies have investigated the role of varicocele repair in patients with NOA, looking particularly at the chances of retrieval of sperm in the ejaculate post-surgery. A meta-analysis done by Esteves et al, evaluated the impact of varicocelectomy on patients with clinically significant varicocele and NOA and showed increased sperm retrieval rate on operated patients[26].

Additionally, in a meta-analysis by Chen et al., varicocelectomy was shown to significantly improve post-operative testosterone levels by 34 ng/dL. In a subgroup analysis of hypogonadal men who underwent varicocelectomy, patients showed a testosterone increase by 105.65 ng/dL as compared to their non-operated counterparts[27,28]. Despite the suggested benefit of varicocelectomy in hypogonadal men, patients undergoing surgery for infertility are to be counseled that the benefit of this procedure to treat hypogonadism is yet to be properly determined by prospective randomized controlled trials.

The benefit of varicocelectomy also extends to patients with the most severe forms of oligospermia, potentially allowing for the downgrading of assisted reproductive techniques (ART) required by the couple. A study done by Cayan et al., showed that patients undergoing varicocelectomy have the chance to perform less invasive procedures of assisted reproductive techniques than their non-operated counterparts[29]. In another study, Samplaski and colleagues showed that men undergoing varicocelectomy have a statistically significant increase in total motile sperm count with the most pronounced increase in patients with severe oligospermia. This allowed patients to be candidates for IUI or spontaneous pregnancy instead of in vitro fertilization (IVF) or introcytoplasmic sperm injection [30], further demonstrating the value of varicocelectomy in different clinical scenarios, including downgrading the level of invasiveness of ART.

Several studies have attempted to describe predictive factors of successful varicocelectomy, including improvements in sperm density, motility and morphology in addition to pain resolution. A study done by Chen et al., found that high testicular volume (> 29.6 mL), lower serum concentration of FSH (< 11.3), and high number of ligated veins (> 9) are predictors of successful varicocelectomy[31]. Another study done by Altunoluk et al., examined the associated resolution of pain post varicocelectomy and the duration of the pain pre-operatively. It was shown that a duration less than 3 months predicted a better pain resolution in patients undergoing surgery[32]. A nomogram designed by Maimaitiming et al., contained the following variables: age, semen parameters pre-operatively, serum testosterone, and testicular atrophy index. This model was able to predict improvement in sperm count, concentration, and vitality with area under the curve (AUC) of 0.915, 0.986, and 0.924, respectively. The model showed a good predictive performance; however, due to the single-center sample size, the study lacks external validity[19]. Another prediction model was developed by Kandevani and colleagues, showing a predictive value of body mass index, neutrophil/lymphocyte ratio and baseline total motile sperm concentration in predicting successful outcomes of varicocelectomy. A score was given for each of these parameters, the summation of which if greater than 5 would predict varicocelectomy failure with 87.5% sensitivity and 84.6% specificity[20]. This stresses on the importance of the pre-operative variables mentioned, which would significantly impact the outcomes of varicocelectomy and should be “plugged in” the nomograms suggested pre-operatively to estimate the likelihood of success of the surgery.

5 AI IN MEDICINE AND UROLOGY

Urology has been at the forefront of integrating advances in artificial intelligence into its everyday practice. Both the physical and virtual branches of AI have seen a vast application in the field. Physical AI represented by robotic surgery is now widely used by urologists for procedures such as prostatectomy, wherein robotic prostatectomy is gradually replacing both laparoscopic and open prostatectomies. In parallel, methods of integrating machine learning and virtual AI into the everyday life of practicing urologists are also being advanced. Several novel machine learning methods to more efficiently predict prostate cancer have been published, taking for instance a study by Nitta et al. that created models and tested their performance against previously used PSA density and PSA velocity[33]. Deep learning approaches are also being applied for the accurate segmentation of images in fusion biopsy, used to correlate magnetic resonance (MR) and trans-rectal ultrasound (TRUS) imaging, with a similarity coefficient of 93%[34].

An increasing number of studies are proposing artificial intelligence models in the field of andrology and male fertility. A study by GhoshRoy et al. applied and compared 7 industry-standard machine learning models in terms of performance for predicting male infertility[35]. Multiple models were created including support vector machine, random forest, decision tree, naïve bayes, logistic regression, adaboost, and multi-layer perception[35]. The input to the models consisted of data with information on environmental and lifestyle factors considered by the World Health Organization (WHO) to affect male fertility, including the season in which the analysis was performed, the age at time of analysis, childhood diseases, previous trauma, surgical history, high fever in the last year, smoking, alcohol consumption, and sitting hours per day[35]. The robustness of each model was assessed through cross-validation techniques, with the best model shown to be the random forest model, reaching an accuracy of 90.47% and an AUC of 99.98% in detecting male infertility[35]. The study then goes one step forward by applying Shapley additive explanation (SHAP) to examine the significance and impact of the different factors on each model to transparently and efficiently allow clinicians to understand and validate the results of the prediction given by the models[35]. AI models with potential clinical use in urology have also been developed to classify sperm head characteristics, classify and predict sperm motility, measure sperm concentration, rank sperm cells based on DNA quality, assess sperm morphology, and predict whether a semen sample is suitable for artificial insemination based on videos[36].

Furthermore, in a study by Bachelot et al., machine learning-based models have been applied to predict the success of testicular sperm extraction in patients with NOA, with a random forest model reaching a performance area under the curve (AUC) of 90% with a sensitivity of 100% and specificity of 69.2%, demonstrating a superior performance vis-a-vis multivariate logistic regression models[37]. Such models further strengthen the case for integrating machine learning to urology.

6 AI IN VARICOCELECTOMY

This section discusses the findings of the five articles retrieved. Following an overview of these findings summarized in Table 1, a detailed discussion on the implications and insights from the results will be presented.

In 2016, a systemic review was published by Esteves et al. addressing the benefit of varicocele correction in improving outcomes of the microsurgical testicular extraction of sperm (mTESE) procedure in patients with azoospermia, and suggesting a substantial role for machine learning in predicting fertility in these patients[26]. A basic science study published in 2018 by Perruzza et al. develops an artificial neural network to predict varicocele impact on male fertility in rat models using endocannabinoid gene expression as the only input. The model's prediction has a substantially low average prediction error of only 1%[38]. In 2022, Ory et al. published a study that proposes a machine learning model to predict the prognosis in terms of fertility after the varicocelectomy procedure, based on clinical, hormonal, and laboratory input. The model performed well with an AUC of 0.72 in predicting clinically meaningful upgrading of fertility[39]. Later in 2024, Crafa et al. published a narrative review for predictors of improvement in sperm parameters after varicocelectomy. The study suggests a potential synthetic role for artificial intelligence through the creation of a model that would predict who would potentially respond to treatment based on inputs such as physical examination, scrotal ultrasound, hormonal values, pre-operative sperm parameters, and other markers. Such a model would decrease the rate of unnecessary surgery[40]. Finally, Calogero et al. published a narrative review in 2024 that summarizes the available literature on the intersection of AI and andrology, tackling many topics such as infertility diagnosis, andrological diagnosis, assisted reproductive techniques, imaging, and surgery. This review mentions the study by Ory et al. tackling AI and varicoceles to illustrate the interaction of machine learning and andrology[41].

6.1 Machine learning models versus Nomograms

A nomogram is a two-dimensional diagram or graphical representation with certain variables on one side and an outcome to predict on the other, allowing through simple geometrical construction to prognosticate a clinical event[18]. Machine learning, on the other side, can handle more complex relationships while nomograms can only predict linear ones[18]. Nomograms are currently widely used to make clinical decisions in urology.

A nomogram to predict changes in semen parameters after varicocele repair in infertile men was proposed in 2014 by Samplaski et al.[42]. However, a multi-center study by Jang et al., that tested the external validity of this nomogram, revealed a low explanatory power, challenging its usefulness in the preoperative setting[43]. A more recent study by Ory et al. looked into creating a new model using a machine learning model to predict fertility post-varicocelectomy, and demonstrating its superiority compared to traditional nomograms[39]. This study was peculiar in that it did not focus on a change in semen parameters to define a successful outcome[39]. Such an outcome would overestimate the success of varicocelectomy as not all changes in sperm parameters are clinically significant. Instead, the study took “clinically meaningful” change in sperm parameters as an outcome, defined as granting the patient access to a previously unavailable form of reproduction[30]. For instance, being able to conceive through IUI post-varicocelectomy when only IVF was possible prior to the procedure was considered a clinically meaningful change. The AI model developed was able to predict the outcome with a performance AUC of 0.72[39], performing extremely better than the nomogram previously used, and reaching a performance comparable to nomograms currently used as the standard of care in urology: The Memorial Sloan Kettering prostatectomy nomogram, which predicts the extent of prostate cancer and long-term results following prostatectomy, having an AUC of 0.74[39].

Interestingly, the machine learning model developed by Ory et al. was based on the following inputs: surgical laterality (unilateral or bilateral), baseline semen concentration and FSH[39]. The nomogram on the other hand, included inputs such as varicocele grade, mean sperm concentration, mean volume, motility, age, and others, while still performing poorer as compared to the AI model, despite the higher number of inputs[42].

Although success rates of varicocelectomies vary between 66% and 70%, a change in sperm parameters does not necessarily mean improved fertility[39]. In the study previously mentioned by Cayan et al., less than 50% of patients with oligospermia who were candidates for IUI had a clinically meaningful change post varicocelectomy, which encourages the use and implementation of machine learning models to help us predict the optimal surgical candidate, while simultaneously avoiding unnecessary surgeries and their associated morbidities on patients who might not benefit from them, especially if other means of conception are still available. Therefore, a patient with oligospermia who is expected to undergo IUI and is unlikely to achieve a new means of conception following varicocelectomy may not be advised to undergo the procedure. Such findings encourage the implementation of deep learning models in the current guidelines, to optimize patient care.

However, a study by Esteves et al. demonstrated a benefit of varicocelectomy in improving sperm retrieval rate in patients with azoospermia originally planned to undergo mTESE. This study suggests the need for implementation of a separate machine learning model to predict fertility in patients undergoing mTESE[26].

6.2 Basic research with clinical promises

Further broadening the scope and application of AI, the search for the molecular contributors to fertility is a major topic in basic research, with the endocannabinoid system being recognized as a key player in the field of reproduction, with involvement in several mechanisms related to germ cell and gonadal cell physiology[38]. Taking AI in predicting fertility one step further, an animal study by Perruzza et al. applied an artificial neural network to predict the fertility of rats with varicocele using endocannabinoid gene expression as the only input, achieving an accuracy of 99%[38]. Given that the endocannabinoid system was demonstrated to be conversed across rats and humans[38], this study and others lay the groundwork for the potential of a molecular laboratory test that helps predict fertility with high accuracy in the presence of varicocele in humans.

6.3 Future perspectives

In light of this review, our future perspective on the role of virtual artificial intelligence in the management of varicoceles is its potential integration into our daily guidelines. The American Urological Association advises an elaborate algorithm for varicocele repair depending on clinical factors, semen analysis, and personal preferences[44]. Potential modifications to these guidelines could include an intermediary step involving the use of a deep learning model to predict fertility based on clinical, hormonal, and laboratory markers, and only offering varicocelectomy in case a clinically significant improvement was predicted. Further regular follow-up and model running could also allow the clinician to reconsider the option of surgery in patients who did not undergo a correction on their first visit, if a clinically significant potential improvement is detected. Such modifications are only suggestions that need further observational research, prospective trials, along with elaborate validation, to see the light. On the long run, such modifications would allow a significant decrease in overtreatment and patient dissatisfaction. However, limitations related to the use of AI are concerns for both transparency and patient confidentiality. The storage of data fed into artificial intelligence and the potential for breaches in patient confidentiality are still unclear[41]. Moreover, deep learning models act as black box models, as compared to the transparency of running nomograms where the path from predictors to outcomes can be visually tracked[18]. The concept of “black box” models limits model interpretability and refinement through human efforts, but also comparability to choose the best AI model out of the ones developed that is to be considered standard of management. Virtual AI also needs large hardware and infrastructure that may not be available in all clinical settings to make decisions, limiting patient accessibility[41].

6.4 Limitations

This narrative review has several limitations, starting with the limited number of articles on AI and varicoceles retrieved from the literature. AI has only recently become a prominent topic, making this limitation in available articles expected. However, the number of publications on this topic is anticipated to rise exponentially in the coming years. The studies retrieved used smaller-scale datasets to develop and validate models, limiting generalizability[39]. Two of the five included articles are reviews that briefly looked at AI in the context of varicocelectomy but were still included for completeness[26,40]. No clinical trials were discussed in this review, as none were retrieved from the literature. Moreover, the available literature could be subject to publication bias, as AI research is currently considered cutting-edge, and negative findings could have a harder time being published. The study was also limited in the risk of bias assessment of the retrieved literature.

7 CONCLUSION

Virtual artificial intelligence is slowly infiltrating our daily lives, and will soon be part of every aspect of human activity, creating previously unimaginable changes to the way we navigate life. This intelligence is currently spreading and its application to all different fields is currently being thoroughly investigated. In the fields of urology and andrology, studies using deep learning models have helped predict sperm parameters, prostate cancer, and the success of testicular sperm extraction. Machine learning models have been developed to predict the success of varicocelectomy using sperm parameters, with a performance superior to that of traditional nomograms. Such models need more validation and research, but are anticipated to surpass the accuracy of all current resources, setting forward a new era of varicocele workup and management in the years to come.

References

[1] Leslie SW, Sajjad H, Siref LE. Varicocele. StatPearls. Treasure Island (FL): StatPearls Publishing Copyright © 2024, StatPearls Publishing LLC; 2024.

[2] Alsaikhan B, Alrabeeah K, Delouya G, Zini A. Epidemiology of varicocele. Asian J Androl. 2016;18(2):179–181.

[3] World Health Organization. The influence of varicocele on parameters of fertility in a large group of men presenting to infertility clinics. Fertil Steril. 1992;57(6):1289–1293.

[4] Gorelick JI, Goldstein M. Loss of fertility in men with varicocele. Fertil Steril. 1993;59(3):613–616.

[5] Baazeem A, Belzile E, Ciampi A, Dohle G, Jarvi K, Salonia A, et al. Varicocele and male factor infertility treatment: a new meta-analysis and review of the role of varicocele repair. Eur Urol. 2011;60(4):796–808.

[6] Hu W, Zhou PH, Zhang XB, Xu CG, Wang W. Roles of adrenomedullin and hypoxia-inducible factor 1 alpha in patients with varicocele. Andrologia. 2015;47(8):951–957.

[7] Eisenberg M, Lipshultz L. Varicocele-induced infertility: newer insights into its pathophysiology. Indian J Urol. 2011;27(1):58–64.

[8] Rajfer J, Turner TT, Rivera F, Howards SS, Sikka SC. Inhibition of testicular testosterone biosynthesis following experimental varicocele in rats. Biol Reprod. 1987;36(4):933–937.

[9] Fujisawa M, Yoshida S, Kojima K, Kamidono S. Biochemical changes in testicular varicocele. Arch Androl. 1989;22(2):149–159.

[10] Stahl P, Schlegel PN. Standardization and documentation of varicocele evaluation. Curr Opin Urol. 2011;21(6):500–505.

[11] Dubin L, Amelar RD. Varicocele size and results of varicocelectomy in selected subfertile men with varicocele. Fertil Steril. 1970;21(8):606–609.

[12] Sakamoto H, Saito K, Oohta M, Inoue K, Ogawa Y, Yoshida H. Testicular volume measurement: comparison of ultrasonography, orchidometry, and water displacement. Urology. 2007;69(1):152–157.

[13] Schlegel PN, Sigman M, Collura B, De Jonge CJ, Eisenberg ML, Lamb DJ, et al. Diagnosis and treatment of infertility in men: AUA/ASRM guideline part II. Fertil Steril. 2021;115(1):62–69.

[14] Sung JJY. Introduction to artificial intelligence in medicine. Singapore Med J. 2024;65(3):132.

[15] Hamet P, Tremblay J. Artificial intelligence in medicine. Metabolism. 2017;69S:S36–S40.

[16] Fatima A, Shafique MA, Alam K, Fadlalla Ahmed TK, Mustafa MS. ChatGPT in medicine: A cross-disciplinary systematic review of ChatGPT's (artificial intelligence) role in research, clinical practice, education, and patient interaction. Medicine. 2024;103(32):e39250.

[17] Gon Park S, Park J, Rock Choi H, Ho Lee J, Tae Cho S, Goo Lee Y, et al. Deep learning model for real‑time semantic segmentation during intraoperative robotic prostatectomy. Eur Urol Open Sci. 2024;62:47–53.

[18] Alabi RO, Mäkitie AA, Pirinen M, Elmusrati M, Leivo I, Almangush A. Comparison of nomogram with machine learning techniques for prediction of overall survival in patients with tongue cancer. Int J Med Inform. 2021;145:104313.

[19] Maimaitiming A, Muhemaiti A, Mulati Y, Li X. Nomograms for predicting postoperative sperm improvements in varicocele patients. Eur Urol Open Sci. 2024;59:40–48.

[20] Kandevani NY, Namdari F, Hamidi M, Dialameh H, Behzadi A. Developing a novel prediction model for the impact of varicocelectomy on postoperative fertility. Eur J Transl Myol. 2022;32(2):10411.

[21] EAU Guidelines. Presented at the EAU Annual Congress, Paris. 2024.

[22] Persad E, O'loughlin CA, Kaur S, Wagner G, Matyas N, Hassler-Di Fratta MR, et al. Surgical or radiological treatment for varicoceles in subfertile men. Cochrane Database Syst Rev. 2021;4(4):000479.

[23] Wang J, Xia SJ, Liu ZH, Tao L, Ge JF, Xu CM, et al. Inguinal and subinguinal micro-varicocelectomy, the optimal surgical management of varicocele: a meta-analysis. Asian J Androl. 2015;17(1):74–80.

[24] Bryniarski P, Taborowski P, Rajwa P, Kaletka Z, Życzkowski M, Paradysz A. The comparison of laparoscopic and microsurgical varicocoelectomy in infertile men with varicocoele on paternity rate 12 months after surgery: a prospective randomized controlled trial. Andrology. 2017;5(3):445–450.

[25] Agarwal A, Cannarella R, Saleh R, Boitrelle F, Gül M, Toprak T, et al. Impact of varicocele repair on semen parameters in infertile men: a systematic review and meta-analysis. World J Men's Health. 2023;41(2):289–310.

[26] Esteves S, Miyaoka R, Roque M, Agarwal A. Outcome of varicocele repair in men with nonobstructive azoospermia: systematic review and meta-analysis. Asian J Androl. 2016;18(2):246–253.

[27] Chen X, Yang D, Lin G, Bao J, Wang J, Tan W. Efficacy of varicocelectomy in the treatment of hypogonadism in subfertile males with clinical varicocele: a meta-analysis. Andrologia. 2017;49(10):e12778.

[28] Çayan S, Akbay E, Saylam B, Kadıoğlu A. Effect of varicocele and its treatment on testosterone in hypogonadal men with varicocele: review of the literature. Balkan Med J. 2020;37(3):121–124.

[29] Çayan S, Erdemir F, Özbey ISA, Turek PJ, Kadioğlu A, Tellaloğlu S. Can varicocelectomy significantly change the way couples use assisted reproductive technologies? J Urol. 2002;167(4):1749–1752.

[30] Samplaski MK, Lo KC, Grober ED, Zini A, Jarvi KA. Varicocelectomy to “upgrade” semen quality to allow couples to use less invasive forms of assisted reproductive technology. Fertil Steril. 2017;108(4):609–612.

[31] Chen SS, Chen LK. Predictive factors of successful varicocelectomy in infertile patients. Urol Int. 2011;86(3):320–324.

[32] Altunoluk B, Soylemez H, Efe E, Malkoc O. Duration of preoperative scrotal pain may predict the success of microsurgical varicocelectomy. International Braz J Urol. 2010;36(1):55–59.

[33] Nitta S, Tsutsumi M, Sakka S, Endo T, Hashimoto K, Hasegawa M, et al. Machine learning methods can more efficiently predict prostate cancer compared with prostate-specific antigen density and prostate-specific antigen velocity. Prostate Int. 2019;7(3):114–118.

[34] Goldenberg SL, Nir G, Salcudean SE. A new era: artificial intelligence and machine learning in prostate cancer. Nat Rev Urol. 2019;16(7):391–403.

[35] GhoshRoy D, Alvi P, Santosh K. Unboxing industry-standard AI models for male fertility prediction with SHAP. Healthcare. 2023;11(7):929.

[36] Riegler MA, Stensen MH, Witczak O, Andersen JM, Hicks SA, Hammer HL, et al. Artificial intelligence in the fertility clinic: status, pitfalls and possibilities. Hum Reprod. 2021;36(9):2429–2442.

[37] Bachelot G, Dhombres F, Sermondade N, Haj Hamid R, Berthaut I, Frydman V, et al. A machine learning approach for the prediction of testicular sperm extraction in nonobstructive azoospermia: algorithm development and validation study. J Med Internet Res. 2023;25:e44047.

[38] Perruzza D, Bernabò N, Rapino C, Valbonetti L, Falanga I, Russo V, et al. Artificial neural network to predict varicocele impact on male fertility through testicular endocannabinoid gene expression profiles. BioMed Res Int. 2018;2018:3591086.

[39] Ory J, Tradewell MB, Blankstein U, Lima TF, Nackeeran S, Gonzalez DC, et al. Artificial intelligence based machine learning models predict sperm parameter upgrading after varicocele repair: a multi-institutional analysis. World J Men's Health. 2022;40(4):618–626.

[40] Crafa A, Cannarella R, Condorelli RA, Mongioì LM, Vignera SL, Calogero AE. Predictive parameters of the efficacy of varicocele repair: a review. Asian J Androl. 2024;26(5):441–450.

[41] Calogero AE, Crafa A, Cannarella R, Saleh R, Shah R, Agarwal A. Artificial intelligence in andrology—fact or fiction: essential takeaway for busy clinicians. Asian J Androl. 2024;26(6):600–604.

[42] Samplaski MK, Yu C, Kattan MW, Lo KC, Grober ED, Zini A, et al. Nomograms for predicting changes in semen parameters in infertile men after varicocele repair. Fertil Steril. 2014;102(1):68–74.

[43] Jang WS, Kim KH, Lim KT, Lee J, Heo JE, Kwon H, et al. External validation of the post-varicocele repair semen analysis nomogram to predict total motile sperm count: a multicenter study. Andrologia. 2020;52(11):e13809.

[44] Brannigan RE, Hermanson L, Kaczmarek J, Kim SK, Kirkby E, Tanrikut C. Updates to male infertility: AUA/ASRM guideline (2024). J Urol. 2024;212:789–799.

RIGHTS & PERMISSIONS

2025 The Author(s). UroPrecision published by John Wiley & Sons Australia, Ltd on behalf of Higher Education Press.