Establishment of Machine Learning Models for Estimating the Collagen Content of Uterine Leiomyomas and Prediction of the Effect of Gonadotropin-Releasing Hormone Analog

建立机器学习模型以估计子宫肌瘤胶原蛋白含量并预测促性腺激素释放激素类似物的作用

阅读:1

Abstract

PURPOSE: To establish machine learning models using magnetic resonance imaging (MRI) data for estimating the collagen content of uterine leiomyomas and predicting the drug therapy effects. METHODS: For surgical patients (90 uterine leiomyomas), tumor signal intensity was quantified by multiple MRI sequences before surgery, and collagen content was quantified by trichrome staining after surgery. These results were used to establish prediction models for estimating collagen content using support vector regression (SVR) and ridge regression (Ridge). For patients who received gonadotropin-releasing hormone analogs (41 uterine leiomyomas), MRI was performed before and after treatment. Correlation between the collagen content estimated by prediction models and tumor reduction rate by GnRHa treatment was investigated. RESULTS: SVR and ridge models were able to estimate the collagen content with high accuracy [R (2): 0.579 (95% CI: 0.33-0.66) and 0.570 (0.27-0.62) in cross validation and 0.648 (0.31-0.85) and 0.675 (0.34-0.86) in Internal Validation]. Significant negative correlations [R: -0.714 (-0.58 to -0.81) and -0.700 (-0.56 to -0.80)] were shown between the estimated collagen content and the tumor reduction rate. CONCLUSIONS: Collagen content of uterine leiomyomas can be estimated by machine learning models using MRI data and can predict the effect of drug therapy on tumor reduction.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。