Ali Utku Öz

Prediction of surgery type for uterine fibroids using machine learning algorithms and hormone values

This study aimed to develop and externally validate machine learning (ML)-based models to characterize surgical classification patterns between hysterectomy and myomectomy using fibroid characteristics and female sex hormone profiles. This multicenter study included 600 women with uterine fibroids (UFs) who presented to 3 hospitals. Of these, 362 (60.3%) underwent hysterectomy, while 238 (39.7%) underwent myomectomy. Statistical analyses and ML models were applied to both groups. ML model development was performed using individual and combined inputs of female sex hormones together with fibroid characteristics. Five ML classification algorithms were evaluated, including support vector machines, decision trees, random forests, k-nearest neighbors, and logistic regression. In total, 2555 model–input combinations were tested. The performance of the selected best-performing model was further evaluated using an independent, blinded external validation cohort comprising 30 cases. Women in the hysterectomy group had significantly higher mean age, follicle-stimulating hormone, luteinizing hormone, UF number, UF volume, uterine volume, disease duration, gravidity, parity, and prolactin (PRL) levels compared with the myomectomy group (all P  < .001). In contrast, estradiol and anti-Müllerian hormone levels were significantly lower in the hysterectomy group ( P  < .001). Across all modeling experiments, 2012 of 2555 model–input combinations achieved perfect classification performance (accuracy = 100%) when sex hormone profiles and UF characteristics were jointly used as inputs. Models using UF number alone also demonstrated high predictive performance, with accuracy reaching up to 96%. Agreement between algorithmic predictions and final surgical decisions was observed in 97% of cases, with one discordant case identified at a clinically borderline threshold. ML models trained on hormone profiles and fibroid characteristics were able to reproduce prevailing surgical classification patterns, largely reflecting strong baseline separability driven by age- and menopause-associated hormonal profiles, with consistent performance observed in an independent blinded validation cohort. These findings support the feasibility of quantitatively modeling routine decision structures, while highlighting the need for further validation in clinically heterogeneous and ambiguous cases.

An AI-Driven Clinical Decision Support Framework Utilizing Female Sex Hormone Parameters for Surgical Decision Guidance in Uterine Fibroid Management

Background and Objective: Changes in female sex hormone levels are closely linked to the development and progression of uterine fibroids (UFs). Clinical approaches to fibroid management vary according to guidelines and depend on patient symptoms, fibroid size, and clinician judgment. Despite available diagnostic tools, surgical decisions remain largely subjective. With the advancement of artificial intelligence (AI) and clinical decision support technologies, clinical experience can now be transferred into data-driven computational models trained with hormone-based parameters. To develop a clinical decision support algorithm that predicts surgical necessity for uterine fibroids by integrating fibroid characteristics and female sex hormone levels. Methods: This multicenter study included 618 women with UFs who presented to three hospitals; 238 underwent surgery. Statistical analyses and artificial intelligence-based modeling were performed to compare surgical and non-surgical groups. Training was conducted with each hormone—follicle-stimulating hormone (FSH), luteinizing hormone (LH), estrogen (E2), prolactin (PRL), and anti-Müllerian hormone (AMH)—and with 126 input combinations including hormonal and morphological variables. Five supervised learning algorithms—support vector machine, decision tree, random forest, and k-nearest neighbors—were applied, resulting in 630 trained models. In addition to this retrospective development phase, a prospective validation was conducted in which 20 independent clinical cases were evaluated in real time by a gynecologist blinded to both the model predictions and the surgical outcomes. Agreement between the clinician’s assessments and the model outputs was measured. Results: FSH, LH, and PRL levels were significantly lower in the surgery group (p < 0.001, 0.009, and <0.001, respectively), while E2 and AMH were higher (p = 0.012 and 0.001). Fibroid volume was also greater among surgical cases (90.8 cc vs. 73.1 cc, p < 0.001). The random forest model using LH, FSH, E2, and AMH achieved the highest accuracy of 91 percent. In the external validation phase, the model’s predictions matched the blinded gynecologist’s decisions in 18 of 20 cases, corresponding to a 90% concordance rate. The two discordant cases were later identified as borderline scenarios with clinically ambiguous surgical indications. Conclusions: The decision support algorithm integrating hormonal and fibroid parameters offers an objective and data-driven approach to predicting surgical necessity in women with UFs. Beyond its strong internal performance metrics, the model demonstrated a high level of clinical concordance during external validation, achieving a 90% agreement rate with an independent, blinded gynecologist. This alignment underscores the model’s practical reliability and its potential to reduce subjective variability in surgical decision-making. By providing a reproducible and clinically consistent framework, the proposed AI-based system represents a meaningful advancement toward the validated integration of computational decision tools into routine gynecological practice.