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. Conclusion: 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.

Uterine fibroids (UFs) are benign monoclonal neoplasms originating from the myometrium and represent the most prevalent tumors in women worldwide. [1] The clinical manifestations of UFs commonly include abnormal uterine bleeding leading to anemia, fatigue, chronic vaginal discharge, and dysmenorrhea. Also referred to as leiomyomas, UFs are among the most common benign neoplasms in women. Reported incidence rates vary widely, ranging from 5.4% to 77%, depending on the studied population. [2,3] Treatment options for UFs include nonsteroidal anti-inflammatory drugs, vitamin D3 and iron supplementation, combined oral contraceptives, gonadotropin-releasing hormone (GnRH) analogs, and surgical excision. [4,5] Female sex hormones play a crucial role in fibroid growth and proliferation. This hormonal dependency allows the use of hormone therapy to reduce fibroid size. Hormone-based treatments may be utilized to alleviate symptoms or facilitate preoperative preparation. GnRH agonists inhibit fibroid growth, thereby reducing menstrual bleeding and pain. In cases of menorrhagia-induced anemia, this treatment can also improve hematologic parameters. However, not all patients respond favorably to these hormonal agents; approximately half report minimal or no symptomatic improvement. The suitability of GnRH agonists depends on fibroid type and the planned surgical approach. Several studies have demonstrated their efficacy as preoperative agents prior to UF surgery.[6] Furthermore, GnRH antagonists have recently emerged as promising alternatives for UF management. These agents rapidly bind to GnRH receptors, block endogenous GnRH activity, and directly suppress LH and FSH secretion, thereby avoiding the initial flare-up effect. [7]Clinical management of UFs may vary among specialists depending on existing guidelines, patient symptoms, and clinical findings. With the rapid advancement of artificial intelligence (AI) and clinical decision support systems, these decision-making processes can now be modeled algorithmically, enabling the translation of clinical experience into computational systems.[8] Such systems can assist less-experienced clinicians in making informed and consistent surgical decisions. A critical objective is to develop algorithms that achieve high predictive accuracy. Moreover, designing AI algorithms that operate with minimal input parameters enhances their practicality for routine clinical use. Vetrivel et al.[9] developed a machine learning–based decision support tool for UF treatment. Using data sourced from Kaggle, the authors trained multiple machine learning (ML) models to predict both treatment decisions and timing. The highest model accuracy achieved was 78%. The present study aims to develop a machine learning–based clinical decision support algorithm to guide surgical decision-making for UFs using fibroid characteristics and female sex hormone parameters.

Ethical Consideration

Ethical approval was granted by the Istinye University Ethics Committee (approval date: June 30, 2025; decision number: 24-18). All study procedures adhered to the ethical standards outlined in the Declaration of Helsinki.

Patients

 A total of 618 patients diagnosed with UFs who presented to the Departments of Obstetrics and Gynecology at Private Derindere Hospital, Ataköy Medicana Hospital, and Kızılay Kağıthane Hospital (Turkey) were included in the study. Of these, 238 patients underwent surgical intervention. Analyses were planned and conducted by stratifying patients into surgical and non-surgical groups.

Study Design

This study is a national, multicenter, retrospective analysis. Statistical analyses were conducted to examine the relationships between surgical and non-surgical groups across all input parameters. ML training was initially performed for each female sex hormone—follicle-stimulating hormone (FSH), luteinizing hormone (LH), estrogen (E2), prolactin (PRL), and anti-Müllerian hormone (AMH)—independently of UF characteristics. Subsequently, 126 unique input combinations were generated by integrating hormone parameters with UF characteristics, and ML models were trained accordingly. In addition to this retrospective development phase, the model underwent a prospective validation process, during which independent clinical cases were evaluated in real time by a gynecologist blinded to model predictions.

Validation Procedure

To evaluate the external decision-support performance of the developed AI model, an independent validation exercise was conducted using 20 anonymized clinical cases that were not included in the training dataset. These cases were presented to an experienced gynecologist who was blinded to both the model predictions and the surgical outcomes. The clinician was asked to assess each case solely based on the available clinical and hormonal parameters and to determine whether myomectomy was indicated.

Statistical Analyses and Tools

Statistical analyses and ML model training were conducted using Wistats v3.0 (WisdomEra Corp., Istanbul, Turkey), incorporating Python-based libraries (SciPy, scikit-learn, statsmodels). Data distribution was assessed via skewness, kurtosis, and the Shapiro–Wilk test. Comparative analyses employed Chi-square or Fisher’s exact tests for categorical variables, and Kruskal–Wallis, one-way ANOVA, t-test, or Mann–Whitney U tests for categorical–numerical comparisons. Pearson and Spearman correlations were used for numerical associations. Multivariate logistic regression and other ML algorithms evaluated predictive performance. A p value < 0>

Machine Learning Procedure and Pipeline

Statistical analyses were conducted prior to ML modeling. Surgical application status was defined as the primary outcome, and predictive algorithms were developed using five classification models: support vector machine (SVM), decision tree (DT), random forest (RF), logistic regression (LR), and k-nearest neighbors (KNN). A 70:30 train–test split was applied for model evaluation, and model performance was evaluated using area under the curve (AUC), accuracy, sensitivity, precision, and F1 score (Figure 1).

Figure 1: Machine Learning Procedure and Pipeline. This figure illustrates the complete workflow used to develop the clinical decision support framework. The process begins with data collection and preprocessing, followed by statistical analyses to identify meaningful hormonal and fibroid-related predictors. Five supervised learning algorithms—random forest, decision tree, logistic regression, support vector machine, and k-nearest neighbors—were trained using multiple input combinations. The model with the highest predictive performance was selected and subsequently deployed on the WisdomEra artificial intelligence platform for real- time clinical decision support. This figure demonstrates the stepwise transformation of raw clinical data into a functional and deployable computational tool. Deploying The Best Decision Support Algorithm on “JinekoAI.com” Web Application Among the developed algorithms, it was determined that the machine learning model with the highest performance could be used in clinical settings. Consequently, the algorithm was implemented on the WisdomEra Artificial Intelligence (WAI) data analytics platform, hosted on cloud infrastructure.[10] In this application, we introduce several decision support algorithms that have been developed, which will be linked to our corresponding publications. Additionally, the algorithms can generate outputs related to the necessity of surgical intervention based on input parameters (Figure 2). This system allows for the integration of the algorithms into decision support services. In doing so, we can transform these decision support algorithms into tangible products for use by both specialists and patients.[11] implemented clinical decision support tool. The interface allows clinicians to input hormone values—including follicle-stimulating.

Figure 2: User Interface of the Deployed Machine Learning Model. This figure presents the graphical user interface of the hormone, luteinizing hormone, estradiol, and anti-Müllerian hormone— and instantly receive individualized surgical guidance. The output section displays the model’s decision, such as whether surgical intervention may be necessary. Model performance indicators, including test dataset proportion, accuracy score, and area under the receiver operating characteristic curve, are also provided. This interface demonstrates how the machine learning model is operationalized into a practical decision-support environment intended to enhance clinical workflow and assist clinicians in real-time patient management.

Figure 3: Comparison of Hormonal and Morphological Parameters Between Surgical and Non-Surgical Groups. This figure presents a grouped bar chart illustrating mean values of key hormonal markers—follicle-stimulating hormone, luteinizing hormone, estrogen, prolactin, and anti-Müllerian hormone—together with fibroid and uterine volumes in women who underwent surgical intervention versus those managed conservatively. For each parameter, black bars represent the surgical group and gray bars represent the non-surgical group. Corresponding p-values above each pair of bars indicate the statistical significance of between-group differences. The chart demonstrates that the surgical cohort exhibited significantly lower follicle-stimulating hormone, luteinizing hormone, and prolactin levels, and significantly higher estrogen and anti-Müllerian hormone levels, as well as larger fibroid and uterine volumes. These trends reflect the combined hormonal and structural features associated with surgical indication in uterine fibroid management.

Statistical Results

The detailed descriptive and comparative statistics are summarized in Table 1. There was no statistically significant age difference between the surgical and non-surgical groups (p = 0.613; mean age: 35.7 vs. 35.4 years). Similarly, the duration of disease (1–5 years vs. > 5 years) showed no statistically significant variation between the two groups (p = 0.361; proportion of patients with disease > 5 years—surgical: 47%, non-surgical: 43%). In contrast, serum levels of FSH, LH, and PRL were significantly lower in the surgical group than in the non-surgical group (p < 0 xss=removed xss=removed>

Table 1:Case Characteristics and Comparative Results Between Surgery and Non-surgery Groups.

Machine Learning Model Training Results

ML models were trained and evaluated to predict surgical necessity by generating 126 input combinations derived from hormonal parameters and UF characteristics. Five different ML algorithms were applied to each input combination, resulting in a total of 630 independent model training sessions. Among these, 12 models achieved an accuracy exceeding 85% and are presented in Table 2. All trained models were categorized according to their accuracy ranges (>90%, 80–90%, 70– 80%, 60–70%, 50–60%, and <50>

Table 2: Machine Learning Models With An Accuracy Score Greater Than 85%.

RF: Random Forest, KNN: K-nearest neighbors, FSH: Follicle stimulating hormone, LH: luteinizing hormone, E2: estrogen, PRL: prolactin, AMH: antimüllerian hormone, UF number: Uterine fibroid number, UF volume: Uterine fibroid volume.

Table 3: Number of Machine Learning Models BasedOn Accuracy Rate Groups.

In multivariable analyses, the most efficient ML model was identified among those utilizing hormonal parameters. Specifically, the RF model trained on LH, FSH, E2, and AMH demonstrated the highest predictive performance, achieving an accuracy of 91%. The test dataset constituted 30% of the total sample. Additional performance metrics were as follows: area under the curve (AUC) = 0.88, Precision = 0.91, Recall = 0.91, and F1-score = 0.91.[12]

When UF volume—a parameter found to differ significantly between surgical and non-surgical groups—was incorporated alongside LH, FSH, E2, and AMH, the resulting model achieved an accuracy of 85%. Using the same 30% test set, its performance metrics were recorded as: AUC = 0.82, Precision = 0.85, Recall = 0.85, and F1-score = 0.85.

External Validation Results

 Concordance was observed in 18 of the 20 cases, corresponding to a 90% agreement rate. The two discordant cases were subsequently reviewed and determined to represent borderline clinical scenarios in which surgical decision-making is inherently ambiguous. This high level of agreement supports the model’s robustness and demonstrates its potential to align closely with expert clinical judgment in real-world decision-making contexts.

UFs substantially affect women’s health and quality of life, primarily through abnormal or heavy menstrual bleeding and iron deficiency anemia.[13] Making timely and appropriate surgical decisions is therefore critical for optimizing patient outcomes. However, determining the optimal timing of surgery often poses a clinical challenge. The management of UFs should be individualized based on symptom severity and the patient’s reproductive preferences or the desire for definitive treatment.[5] Given the well-established role of female sex hormones in UF pathophysiology [14], our findings demonstrate that hormonal parameters can effectively contribute to clinical decision- support algorithms for predicting surgical necessity. Notably, the RF model incorporating LH, FSH, E2, and AMH achieved the highest predictive accuracy (91%). Furthermore, the UF volume was significantly higher in the surgical group compared with the non-surgical group (p < 0>

In conclusion, determining the appropriate timing for UF surgery remains a global clinical challenge. The AI-based clinical decision-support algorithms developed in this study can effectively assist in making timely and evidence-based surgical decisions. The high external validation concordance—90% agreement between the model and a blinded gynecologist—demonstrates that the system not only performs well statistically but also aligns closely with real clinical judgment. Our findings underscore the potential of AI to augment clinical expertise, reduce subjective variability, and optimize management strategies for patients with UFs. Further work is underway to refine and expand these AI models for broader surgical decision-making contexts.

Conceptualization: İnci OZ and Engin Ulukaya; Methodology: İnci OZ and Engin Ulukaya; Investigation: İnci OZ, Ali Utku OZ, and Ecem E. YEGIN; Resources: İnci OZ and Ali Utku OZ; Data Curation: İnci OZ; Writing – Original Draft Preparation: İnci OZ, Ali Utku OZ, and Ecem E. YEGIN; Writing – Review & Editing: İnci OZ and Ecem E. YEGIN; Supervision: İnci OZ and Engin Ulukaya.

This research received no external funding.

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Istinye University (protocol code 24-18 and January 30, 2025).

Patient consent was waived because the study was designed as a retrospective analysis of anonymized medical records, involved no direct patient contact or intervention, and posed no foreseeable risk to participants. The Institutional Review Board approved the waiver of informed consent in accordance with national regulations and the Declaration of Helsinki.

The dataset generated and analyzed in this study is publicly available on the Istinye University Dataset Sharing Platform. Anonymized clinical and hormonal data related to uterine fibroids can be accessed at the following link: https://dataset.istinye.edu.tr/dataset?did=55. All data were fully anonymized in accordance with ethical regulations. Access is provided for research purposes through a controlled-access system under the platform’s standard licensing and data-sharing policies.

We would like to thank the Artificial Intelligence Research And Application Center of Istinye University (https://yzaum.istinye.edu.tr/) for their support in the technical assessment of the manuscript, including verification of data integrity and plagiarism screening prior to submission. We also extend our appreciation to the Ditako Data Analytics Team (https://ditako.com) for providing professional statistical analysis and machine learning services that contributed to the robustness of the results.

None of the authors has any potential financial conflict of interest related to this manuscript.

The following abbreviations are used in this manuscript:

AI                                                                                    Artificial intelligence

AMH                                                                              Anti-Müllerian hormone

AUC                                                                               Area underthe ROC curve

cc                                                                                    Cubic centimeter

DT                                                                                  Decision tree

E2                                                                                   Estradiol

FSH                                                                                Follicle-stimulating hormone

GnRH                                                                             Gonadotropin-releasing hormone

KNN                                                                              k-nearest neighbors

LH                                                                                  Luteinizing hormone

LR                                                                                  Logistic regression

ML                                                                                 Machine learning

PRL                                                                                Prolactin

RF                                                                                  Random forest

SVM                                                                              Support vectormachine

UF                                                                                  Uterine fibroid

UFs                                                                                Uterine fibroids

WAI                                                                                Wisdom Era  Artificial Intelligence