Lavanya J M Sheela, P Subbulakshmi
School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India.
Heliyon. 2024 Apr 15;10(9):e29197. doi: 10.1016/j.heliyon.2024.e29197. eCollection 2024 May 15.
Globally, ovarian cancer affects women disproportionately, causing significant morbidity and mortality rates. The early diagnosis of ovarian cancer is necessary for enhancing patient health and survival rates. This research article explores the utilization of Machine Learning (ML) techniques alongside eXplainable Artificial Intelligence (XAI) methodologies to aid in the early detection of ovarian cancer. ML techniques have recently gained popularity in developing predictive models to detect early-stage ovarian cancer. These predictions are made using XAI in a transparent and understandable way for healthcare professionals and patients. The primary aim of this study is to evaluate the effectiveness of various ovarian cancer prediction methodologies. This includes assessing K Nearest Neighbors, Support Vector Machines, Decision trees, and ensemble learning techniques such as Max Voting, Boosting, Bagging, and Stacking. A dataset of 349 patients with known ovarian cancer status was collected from Kaggle. The dataset included a comprehensive range of clinical features such as age, family history, tumor markers, and imaging characteristics. Preprocessing techniques were applied to enhance input data, including feature scaling and dimensionality reduction. A Minimum Redundancy Maximum Relevance (MRMR) algorithm was used to select the features in the model. Our experimental results demonstrate that in Support Vector Machines, we found 85 % base model accuracy and 89 % accuracy after stacking several ensemble learning techniques. With the help of XAI, complex ML algorithms can be given more profound insights into their decision-making, improving their applicability. This paper aims to introduce the best practices for integrating ML and artificial intelligence in biomarker evaluation. Building and evaluating Shapley values-based classifiers and visualizing results were the focus of our investigation. The study contributes to the field of oncology and women's health by offering a promising approach to the early diagnosis of ovarian cancer.
在全球范围内，卵巢癌对女性的影响尤为严重，导致了较高的发病率和死亡率。卵巢癌的早期诊断对于提高患者健康水平和生存率至关重要。本文探讨了如何利用机器学习（ML）技术以及可解释人工智能（XAI）方法来辅助卵巢癌的早期检测。近年来，ML技术在开发用于检测早期卵巢癌的预测模型方面颇受关注。这些预测通过XAI以一种对医疗专业人员和患者来说透明且易懂的方式进行。本研究的主要目的是评估各种卵巢癌预测方法的有效性。这包括评估K近邻算法、支持向量机、决策树以及集成学习技术，如多数投票、提升、装袋和堆叠。从Kaggle收集了一个包含349名已知卵巢癌状况患者的数据集。该数据集包括一系列全面的临床特征，如年龄、家族病史、肿瘤标志物和影像特征。应用预处理技术来增强输入数据，包括特征缩放和降维。使用最小冗余最大相关性（MRMR）算法在模型中选择特征。我们的实验结果表明，在支持向量机中，我们发现基础模型准确率为85%，在堆叠几种集成学习技术后准确率为89%。借助XAI，可以更深入地了解复杂ML算法的决策过程，提高其适用性。本文旨在介绍在生物标志物评估中整合ML和人工智能的最佳实践。构建和评估基于沙普利值的分类器并可视化结果是我们研究的重点。该研究通过提供一种有前景的卵巢癌早期诊断方法，为肿瘤学和女性健康领域做出了贡献。
Front Public Health. 2025-3-26
BMC Infect Dis. 2025-3-26
Discov Ment Health. 2025-7-1
Discov Oncol. 2025-5-13
Front Public Health. 2025-3-26
Comput Methods Programs Biomed. 2022-2
Lancet Digit Health. 2021-11
Biomed Opt Express. 2021-4-2
Zotero 插件
