(383) Artificial Intelligence in the Diagnosis and Treatment Guidance of Erectile Dysfunction

Introduction Erectile dysfunction (ED) is a multifactorial condition that significantly impairs male quality of life. Diagnosis and treatment often rely on subjective evaluations and a trial-and-error approach in majority of urologic center. Artificial intelligence (AI) can analyze large volumes of clinical data and identify complex patterns, enabling a more accurate and personalized diagnostic and therapeutic process. Objective To develop and validate an AI-based predictive model capable of diagnosing ED and recommending the most effective treatment based on clinical, hormonal, and functional data. Methods A retrospective analysis was conducted involving 542 men aged 25 to 75 years at between 2018 and 2025. Collected data included clinical variables (age, comorbidities, BMI, smoking, medications), hormonal (total testosterone, prolactin, estradiol, glicemia), functional (nocturnal erection self-reports, pharmacologically induced erection tests), psychometric (IIEF-5, anxiety and depression scales), and sexual history. Machine learning models were trained to predict confirmed ED diagnosis (IIEF-5 ≤ 21) and the most effective treatment. XGBoost (AI) was selected as the final model and validated on a 20% holdout sample. Results The results are: diagnostic accuracy of 94.2%, sensitivity of 92,1%, specificity of 95.6%. Most predictive valiables were IIEF-5 score, diabetes levels, Total Testosterone, age of patients, anxiety levels. The treatments recommendations accuray were 88,3%. The PDE5 inhibitors (Tadalafl and Sildenafil) 91,0% ( mild to moderate cases). Hormone therapy 93,0% (hypogonadism). Psychotherapy: 79,0% to psychogenic causes, Combination therapy in 86% to refratory cases. Conclusions AI demonstrated high precision in diagnosing erectile dysfunction and personalizing treatment strategies. By integrating clinical, hormonal, and functional data, the model provides strong decision-making support for clinicians, optimizing therapeutic outcomes and minimizing treatment failures. Disclosure No