Application of Big Data Technology to Assessments of Female Ovarian Reserve Dysfunction

作     者：Xia, Ji'An Ma, Yunfei Wu, Yiyun Zhao, Youlin Ni, Haorang Liu, Xinyan 

作者机构：Nanjing Vocat Univ Ind Technol Sch Comp & Software Nanjing 210023 Jiangsu Peoples R China Jiangsu Prov Hosp Chinese Med Dept Ultrasound Nanjing 210029 Jiangsu Peoples R China 

出 版 物：《IEEE ACCESS》 (IEEE Access)

年 卷 期：2025年第13卷

页      面：19408-19419页

核心收录：

基　　金：Jiangsu Province Public Health and Social Development Project [BE2022789] Jiangsu Industrial Software Engineering Technology Fund [ZK20-04-12] Jiangsu Province 333 High Level Talent Training Project [3-25-069] Introduction of Talent Research Fund [YK20-05-01] 

主　　题：Medical diagnostic imaging Big Data Machine learning algorithms Sparks Classification algorithms Indexes Data models Random forests Biochemistry Support vector machines Diminished ovarian reserve three-dimensional power Doppler ultrasound medical big data machine learning classification algorithm performance evaluation 

摘      要：This paper evaluated the potential application of big data technology to assessments of diminished ovarian reserve (DOR). The study enrolled 162 patients who underwent ovarian reserve function assessment for the first time in the Department of Ultrasound, Jiangsu Province Hospital of Chinese Medicine from January 2023 to December 2023. Patients were divided into normal ovarian reserve function (n = 68), early-stage DOR (n = 66), mid-stage DOR (n = 12), and late-stage DOR (n = 16) groups. Hadoop and Spark frameworks were used to build a big data platform, and the MLlib parallel machine learning library was used to implement three multivariate classification models-multilayer perceptron, one-vs-rest, and random forest classifiers-to classify and analyse the ovarian reserve function dataset and evaluate the platform s performance. In the big data platform, the random forest algorithm achieved the highest classification accuracy (89.47%), followed by the neural network (81.06%) and support vector machine (72.91%) methods. The random forest algorithm had the least time overhead for datasets smaller than 50 MB;for datasets exceeding 50 MB, the support vector machine algorithm had the least time overhead, followed by the random forest and neural network algorithms. The neural network algorithm s speedup ratio was lower than that of the other two algorithms for small datasets, but with increasing dataset size, its speedup ratio significantly exceeded those of the other two algorithms. The random forest algorithm showed substantial growth for large datasets.