Machine-learning diagnostic models for ovarian tumors

Yuwei Sun ¹, Bin Wen ¹

⁰Department of Gynecology, Guangdong Women and Children Hospital, Guangzhou City, Guangdong Province, China.

Heliyon +

|

October 9, 2024

View abstract on PubMed

Summary

Machine learning models can accurately diagnose ovarian tumors using routine blood tests and patient age. These predictive algorithms offer a framework for personalized treatment strategies and improved patient stratification.

Area Of Science

Oncology
Biomedical Informatics
Machine Learning

Background

Ovarian cancer diagnosis and prognosis are critical for effective treatment.
Current diagnostic methods may lack precision in predicting tumor behavior and pathological outcomes.
There is a need for advanced analytical tools to improve diagnostic accuracy and prognostic capabilities.

Purpose Of The Study

To develop a diagnostic framework for ovarian tumors using machine learning (ML).
To predict clinical behavior and pathological tissue prognosis based on multiple biomarkers.
To leverage ML methods for enhanced diagnostic and prognostic information.

Main Methods

Utilized supervised ML classifiers including Support Vector Machine, Random Forest, k-nearest neighbor, and logistic regression.
Trained models on data from 713 ovarian tumor patients, incorporating 10 blood test parameters and age.
Validated models using accuracy, area under the ROC curve, and external patient data.

Main Results

ML techniques outperformed conventional regression analyses in predicting ovarian tumor characteristics.
Random Forest (RF) demonstrated superior diagnostic performance for malignant ovarian cancer, achieving 99.82% accuracy.
Logistic regression showed high accuracy (78.0%) for pathological tissue diagnosis, with external validation yielding 71.9% accuracy.

Conclusions

ML systems can diagnose and characterize ovarian tumors pre-treatment, providing crucial prognostic insights.
Predictive algorithms facilitate personalized treatment planning through patient preprocessing stratification.
This approach enhances diagnostic precision and supports tailored therapeutic strategies for ovarian cancer.

Keywords:

Machine learning Ovarian tumor Tumor markers