可解释的机器学习用于预测慢性病进展风险和预测慢性病进展风险
Jin-Xin Zheng1, Xin Li1, Jiang Zhu2
1Department of Nephrology, Ruijin Hospital, Institute of Nephrology, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
Digital health
|January 18, 2024
概括
可解释的机器学习模型准确地预测慢性病 (CKD) 的进展. 极端梯度提升和随机生存森林提供了卓越的性能和对年龄和肌素等风险因素的洞察力.
科学领域:
- 腎臟病學 (nephrology) 是一種醫學專業.
- 医疗信息学 医疗信息学
- 生物统计学 生物统计学
背景情况:
- 慢性病 (CKD) 是一个重大的全球健康挑战.
- 准确的早期预测CKD进展对于及时干预至关重要.
- 传统的预测模型往往缺乏临床使用所需的准确性和可解释性.
研究的目的:
- 开发和比较用于预测CKD进展的机器学习 (ML) 模型.
- 提高临床决策的ML模型的可解释性.
- 确定关键的风险因素及其与CKD进展的关联.
主要方法:
- 使用了491名具有临床数据的患者队列,随机分为训练和测试组.
- 开发并评估了四个ML算法 (逻辑回归,随机森林,神经网络,XGBoost) 用于分类.
- 使用Cox比例危险回归 (COX) 和随机生存森林 (RSF) 进行生存分析.
- 使用AUC-ROC,C指数和集成的Brier分数来评估模型性能.
- 使用变量重要性,部分依赖图和受限制的立方支线的解释模型.
主要成果:
- 在CKD分类中,XGBoost获得了最高的预测性能 (AUC-ROC:0.867).
- 随机生存森林 (RSF) 在生存分析中表现出优异的歧视和校准能力,与COX.
- 慢性瘤进展的关键预测因素包括估计的膜过率,年龄和肌素.
- 确定了年龄和胆固醇水平与CKD进展之间的非线性关联.
结论:
- 可解释的ML模型对于预测CKD进展是有效的.
- ML,特别是RSF,在CKD的生存分析中比传统方法提供了优势.
- 这项研究通过阐明非线性风险因素关系和比较预测技术来推进CKD风险预测.
更多相关视频
07:15Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.8K
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
8.3K
相关概念视频
Factors Affecting Renal Clearance: Renal Impairment
95
Renal dysfunction significantly impairs the renal clearance of drugs, leading to potential complications in drug therapy. Renal failure, which can be caused by various factors, poses a significant challenge in the elimination of drugs from the body.
One condition associated with renal failure is uremia. Uremia is characterized by impaired glomerular filtration and fluid accumulation in the body. This condition hinders the renal clearance of drugs, resulting in drug accumulation and potential...
One condition associated with renal failure is uremia. Uremia is characterized by impaired glomerular filtration and fluid accumulation in the body. This condition hinders the renal clearance of drugs, resulting in drug accumulation and potential...
95
Dialysis
319
Renal failure occurs when the kidneys lose their ability to filter waste products from the blood effectively. It can be classified into two types: acute renal failure (ARF) and chronic renal failure (CRF).
Acute kidney injury develops suddenly and can be caused by pre-renal causes (e.g., hypovolemia, shock), intrinsic renal causes (e.g., acute tubular necrosis), or post-renal causes (e.g., urinary obstruction). In contrast, chronic renal failure progresses gradually over time and is often...
Acute kidney injury develops suddenly and can be caused by pre-renal causes (e.g., hypovolemia, shock), intrinsic renal causes (e.g., acute tubular necrosis), or post-renal causes (e.g., urinary obstruction). In contrast, chronic renal failure progresses gradually over time and is often...
319
Nephrons
2.5K
The kidneys are intricate organs with millions of working units known as nephrons. Each nephron features two major structures: the renal corpuscle, which facilitates blood plasma filtration, and the renal tubule, which handles the glomerular filtrate. Blood supply is directly linked to the nephrons. The renal corpuscle consists of the glomerulus, a capillary network, and the Bowman's capsule, a double-walled epithelial structure that encases the glomerulus. The filtering of blood plasma...
2.5K
Receiver Operating Characteristic Plot
183
A ROC (Receiver Operating Characteristic) plot is a graphical tool used to assess the performance of a binary classification model by illustrating the trade-off between sensitivity (true positive rate) and specificity (false positive rate). By plotting sensitivity against 1 - specificity across various threshold settings, the ROC curve shows how well the model distinguishes between classes, with a curve closer to the top-left corner indicating a more accurate model. The area under the ROC curve...
183
