Male infertility accounts for nearly half of all cases of couple infertility; however, currently available diagnostic tools remain limited.
Semen analysis, the standard reference method, provides only partial insight into spermatogenic function, while testicular ultrasound (TUS), although an essential diagnostic modality, is largely descriptive in nature and highly operator-dependent.
Our project addresses these limitations by integrating radiomics and artificial intelligence (AI) into testicular ultrasound, with the aim of developing objective, reproducible, and clinically meaningful assessments of testicular function. Preliminary studies, including a large cohort of 302 men and the INTACT pilot study (85 patients, 170 images), have demonstrated that testicular ultrasound texture features correlate with semen parameters and endocrine function, providing robust proof of concept for AI-driven quantification of testicular echotexture.
The primary objective of the project is to evaluate the correlation between ultrasound texture features and spermatogenic capacity. Secondary objectives include investigating the associations between imaging parameters and endocrine function, as well as developing AI-based predictive algorithms for the assessment of testicular function.
Specific Aim 1 involves the retrospective analysis of static and dynamic testicular ultrasound images using radiomic feature extraction and AI modeling, with the goal of reducing operator-dependent variability and standardizing image interpretation.
Specific Aim 2 focuses on the prospective validation of these algorithms in men attending infertility clinics, assessing their accuracy, reproducibility, and clinical utility.
Comprehensive clinical data—including anthropometric parameters, hormonal profiles, semen analysis, genetic testing, and information related to the female partner—will be integrated with imaging data to develop robust AI models capable of predicting spermatogenic potential and response to hormonal stimulation, with the potential to reduce or avoid the need for invasive testicular biopsy.
The study is structured in three phases: retrospective data collection, AI-based image analysis and model development, and prospective validation. Expected outcomes include standardized, operator-independent tools for testicular ultrasound interpretation, capable of linking testicular morphology with functional fertility parameters. By enhancing diagnostic precision, reducing reliance on variable semen analysis, and providing clinically actionable predictive information, this project will support personalized clinical management, optimize healthcare resource utilization, and improve patient outcomes.
This interdisciplinary approach integrates andrology, endocrinology, radiology, and artificial intelligence, introducing a novel paradigm for the assessment of male infertility. Results will be disseminated through high-impact scientific publications, international conferences, and training programs. By delivering innovative, non-invasive, and reliable diagnostic tools, the project addresses a major unmet clinical need, promotes digital innovation in reproductive medicine, and contributes to improved healthcare efficiency and societal well-being.