Artificial Intelligence in Onco-Andrology: Applications in Fertility Preservation, Sperm Analysis and Post-Treatment Recovery

INTRODUCTION

Cancer remains a significant public health burden in the United States, with projections of over two million new cases and more than 600,000 fatalities by 2025.^[1] Despite advancements in diagnosis and predictive treatment, prostate and testicular cancer continue to contribute substantially to global male morbidity and mortality, highlighting their persistent public health significance.^[2] Chemotherapy and radiotherapy are hallmark treatment modalities that have substantially improved survival outcomes over recent decades.^[3] With increasing numbers of male cancer survivors, the need to preserve and restore fertility has emerged as a vital component of comprehensive survivorship care. This shift has given rise to the specialised field of onco-andrology, which bridges oncology and male reproductive health to address the unique fertility challenges faced by cancer patients and survivors.^[4] Although conventional fertility preservation strategies are valuable, their success depends on multiple factors, including semen quality, cancer type, and treatment regimen.^[5,6] These limitations highlight the need for innovative, data-driven approaches capable of providing personalised fertility care.

Artificial Intelligence (AI) has become a revolutionary asset in medical practice and is progressively being integrated into reproductive medicine. Using advanced data-driven algorithms and representation-learning approaches, AI can analyse large, complex datasets to surface subtle patterns, anticipate clinical outcomes and strengthen evidence-based decision-making.^[7] In andrology, AI enables automated analysis of sperm density, motility, and morphology with improved accuracy compared to manual methods. It has also been applied to assess sperm deoxyribonucleic acid (DNA) integrity, predict infertility risk, and guide sperm selection for procedures such as Intracytoplasmic Sperm Injection (ICSI).^[8,9] For a detailed comparison of conventional versus AI-based fertility assessment methods, see Table 1.^[10,11] Despite significant advances, AI in onco-andrology remains fragmented and lacks extensive clinical validation. This review synthesises current evidence, identifies key challenges, and outlines future research directions for integrating AI into fertility care for male cancer survivors.

AI TECHNOLOGIES IN ONCO-ANDROLOGY

AI is swiftly revolutionising onco-andrology by offering advanced solutions that tackle the intricate issues related to fertility preservation and restoration in men affected by cancer. In oncology, AI applications are extensively applied for the early detection, disease forecasting, outcome prediction and treatment planning of diverse malignancies, helping to improve patient outcomes through personalised medicine.^[12,13] In the andrological domain, these technologies are increasingly employed for semen analysis, sperm DNA integrity assessment, infertility risk prediction and optimisation of assisted reproductive techniques (ART).^[14] By leveraging machine learning (ML) and deep learning, AI can integrate clinical, hormonal, genetic and sperm imaging data to detect complex patterns and generate predictive models beyond the scope of traditional analysis.^[15] AI-driven models demonstrate exceptional capability in analysing sperm morphology, achieving classification accuracies of up to 94% and providing consistent evaluations even under varying image qualities. AI-assisted platforms combined with sperm chromatin dispersion tests also demonstrate strong concordance with manual assessments, offering an objective and scalable strategy for DNA fragmentation analysis in sperm selection.^[16,17] Deep learning algorithms further enhance the evaluation of sperm DNA integrity, enabling detection of subtle abnormalities that improve accuracy in predicting infertility risk. In a study by Ghosh Roy et al., seven ML models were compared, and an explainable random forest classifier achieved 90.47% accuracy with an AUC of 0.998 for male fertility prediction, highlighting the potential of interpretable models to provide transparent clinical insights.^[18] Beyond fertility, AI tools are influencing broader aspects of male reproductive oncology. Convolutional Neural Networks have advanced the standardisation of prostate cancer grading by reducing inter-observer variability in Gleason scoring, thereby improving diagnostic consistency.^[19] Furthermore, clinical decision support systems incorporating genetic algorithms and support vector machines have achieved high accuracy in predicting erectile dysfunction based on comprehensive patient datasets, aiding personalised management strategies.^[20] Collectively, AI-powered solutions in onco-andrology facilitate more accurate diagnostic procedures, personalised treatment strategies and enhanced reproductive outcomes for individuals undergoing cancer therapy.

AI IN FERTILITY PRESERVATION AND AUTOMATED SPERM ANALYSIS

Fertility preservation is a critical aspect of cancer care, as chemotherapy and radiotherapy often impair gonadal function and diminish reproductive capacity. For men, sperm cryopreservation remains the foremost option, generally entailing the collection of semen samples prior to initiating cancer therapy to secure biologically viable and genetically stable sperm for potential future use.^[21] ART utilising cryopreserved sperm from cancer survivors have demonstrated meaningful outcomes, with ICSI reporting the highest pregnancy (34%) and delivery rates (23%) compared with In Vitro Fertilisation (IVF) and Intrauterine Insemination.^[5] AI is now being integrated into fertility preservation pathways to enhance clinical decision-making and personalise treatment strategies.^[22] AI-driven decision support systems assist clinicians in formulating personalised fertility preservation strategies that are tailored to the patient’s cancer type, therapeutic history and baseline reproductive profile.^[3,23] Multimodal AI frameworks that combine clinical, hormonal, and imaging data further strengthen assessments, offering a comprehensive view of fertility potential that surpasses traditional single-modality analyses.^[24] Alongside these technological advances, numerous studies have reported a consistent downward trend in male gamete health noted globally in recent generations, underscoring the growing focus on male reproductive health.^[25] In parallel, AI-driven innovations in automated sperm analysis are redefining the accuracy and efficiency of male fertility evaluation. Conventional semen analysis, though standardised by the World Health Organisation, often shows limited predictive value for clinical outcomes.^[26] To overcome these limitations, Urbano et al. developed a fully automated multi-sperm tracking system capable of simultaneously monitoring hundreds of sperm with minimal operator input, overcoming performance issues experienced in conventional computer-assisted sperm analysis systems.^[27] Shahzad et al. proposed a sequential deep neural network that accurately detects sperm morphological abnormalities, including defects in the acrosome and head, achieving accuracies of up to 92% even on unstained, low-resolution images, thereby enabling rapid and clinically useful evaluations.^[28] Kumar and colleagues further applied AI-based deep learning to measure DNA strand breakage in sperm cells using images from the Chromatin Integrity Test, a reliable method for assessing DNA damage, achieving accuracy above 80% and enabling consistent evaluation without the need for expensive software, thus enhancing the clinical evaluation of sperm quality.^[29] Beyond ejaculate-based assessments, AI models have also been extended to testicular biopsy imaging, where deep learning and unsupervised algorithms such as K-means clustering are being utilised to identify sperm cells in patients with azoospermia and investigate the molecular underpinnings of DNA fragmentation.^[30]

PREDICTIVE ANALYTICS FOR POST-TREATMENT FERTILITY RECOVERY

Azoospermia impacts roughly 1% of the male demographic, and although treatments like gonadotropin therapy, varicocelectomy or microdissection testicular sperm extraction may enhance fertility in certain individuals, results are inconsistent due to diverse causes and insufficient evidence.^[31] Meta-analytic data demonstrate that although the majority of men retain fertility following testicular cancer therapy, roughly 14% experience infertility and up to 8% present with azoospermia, emphasising the influence of both treatment modality and individual patient factors on reproductive prognosis.^[32] Recent advances in AI have enabled the development of ML models capable of processing complex, multifactorial datasets to inform optimal treatment selection in ART, thereby enhancing success rates and minimising unnecessary interventions.^[33] Semen cryopreservation is a feasible strategy even in pubertal boys prior to gonadotoxic therapies, underscoring its importance as a proactive measure for fertility preservation despite young age.^[34] Specifically in male infertility, validated AI models such as random forest algorithms have demonstrated robust predictive power in identifying patient subgroups likely to benefit from interventions like varicocelectomy, paving the way for clinical calculators that support more personalised therapeutic decisions.^[35]

To create predictive models for treatment outcomes, Bai R and colleagues applied multiple ML algorithms in a cohort of 2,625 women undergoing fresh IVF embryo transfer. The XGBoost model achieved the highest predictive accuracy for clinical pregnancy (AUC 0.999), while LightGBM performed best for live birth prediction (AUC 0.913), demonstrating the strong potential of AI-based approaches in reproductive medicine.^[36] Although these findings were derived from female-focused infertility care, they provide important proof-of-concept evidence that predictive analytics can generate highly accurate and clinically relevant forecasts. Translating such methodologies to male infertility holds promise for improving individualised treatment selection, guiding fertility-preserving interventions and optimising recovery outcomes after cancer therapy. By harnessing predictive analytics, clinicians can better anticipate post-treatment fertility trajectories, minimise ineffective interventions and deliver more personalised survivorship care.^[37]

ETHICAL, REGULATORY AND CLINICAL ADOPTION CHALLENGES

AI is emerging in reproductive medicine to refine sperm, oocyte, and embryo selection and to strengthen IVF prediction models. It provides new possibilities for couples facing infertility, though its use is still experimental. Ethical concerns remain, as evidence of effectiveness is limited and ensuring proper informed consent is challenging.^[38] Research in reproductive medicine involving humans is especially sensitive from an ethical perspective and poses unique challenges. Initially, the intended mother is the primary participant, carrying most of the physical and psychological burden compared to her male partner. However, studies and innovations in assisted reproduction, including AI applications, ultimately extend their impact to the potential birth of offspring.^[39] AI applications in human reproduction raise profound ethical concerns, given the intimate nature of reproduction and the moral significance of embryos. Addressing both general AI challenges and those unique to this sensitive context is vital to maintaining trust in ML-based embryo assessment and ART. Yet, ethical issues in this area remain underexplored. Key concerns, rooted in broader AI ethics guidelines, centre on accountability, transparency, fairness and respect for autonomy.^[40] It is now widely recognised that AI systems can exhibit unintended biases, which may appear in different forms. One key concern is that ML algorithms may perform better for certain groups than others, for example, across ethnicities, often due to unequal representation of these groups in the training datasets.^[41] In AI ethics and computer science research, the terms ‘transparency’, ‘interpretability’ and ‘explainability’ are often used inconsistently and at times even contradictorily. Interpretable ML models are considered safer and more effective than black-box algorithms in embryo selection, as they enable clinicians to identify confounders, detect flawed reasoning, and adjust to variations across clinical environments. Unlike opaque models, interpretable approaches allow embryologists to investigate and resolve discrepancies, incorporate patient values, and support shared decision-making. By blending ML insights with clinical expertise, responsibility for outcomes stays with the clinician, leading to more accurate, reliable and accountable embryo selection.^[42]

EMERGING DIRECTIONS AND FUTURE PERSPECTIVES

Emerging directions in reproductive medicine highlight the integration of AI with genomics and multi-omics, the growth of digital health in survivorship care, and the application of AI to fertility restoration technologies.^[43] Through multi-omics integration, AI is being used to identify diagnostic markers for polycystic ovary syndrome, endometrial receptivity, and male infertility, as well as to predict ovarian reserve, gamete quality, and IVF outcomes.^[44] Although challenges remain, such as variability across datasets and limited external validation, explainable AI models are improving the transparency and scalability of these applications.^[45] At the same time, digital health solutions, including web-based interventions, mobile monitoring platforms, and electronic survivorship care plans, are transforming long-term care for cancer survivors.^[46] Early studies suggest they enhance symptom tracking, physical activity and psychosocial support, though more robust trials are still needed.^[47] In fertility-specific contexts, digital onco-fertility platforms are improving counselling, decision-making and follow-up care.^[48] Parallel advances in fertility restoration focus on bioprosthetic ovaries and 3D bioprinting technologies, which aim to restore both endocrine and reproductive function while minimising the risk of reintroducing malignant cells. Proof-of-concept studies in animal models have already shown functional restoration.^[49,50] AI plays a key role in these innovations by optimising scaffold design, automating follicle assessment with advanced imaging, and supporting ovarian reserve prediction with tools.^[51] Collectively, these emerging innovations demonstrate how AI is reshaping reproductive medicine by driving discovery, improving survivorship care and enabling next-generation fertility restoration.^[52]

CONCLUSION

AI is transforming the field of onco-andrology through improving the precision of fertility evaluations, guiding personalised treatment strategies and predicting long-term outcomes for male cancer survivors through the integration of genomic insights and digital health technologies. Yet, the advancement of such personalised fertility care requires strong ethical frameworks, strict regulatory measures and fair access to modern diagnostics and therapies. As AI tools evolve alongside interdisciplinary collaboration, healthcare providers must ensure that innovation complements human compassion and transparent communication. Future research must move beyond proof-of-concept models toward clinically validated AI tools. Interdisciplinary collaboration between oncologists, andrologists, data scientists and ethicists will be pivotal in achieving equitable, transparent and effective AI-driven fertility care. When adopted responsibly, these technologies have the potential to improve not only clinical results but also the overall quality of life and reproductive independence of male cancer survivors.