Fed-MStacking:异质联合学习与堆叠错位标签用于异常心脏声音检测
IEEE journal of biomedical and health informatics
|July 16, 2024
概括
本研究介绍了Fed-MStacking,这是一种用于智能医疗保健的新型联合学习框架,增强了心声分析的隐私和性能. 它可以实现个性化,异构的模型,在多机构数据场景中表现优于传统方法.
科学领域:
- 人工智能的人工智能
- 聪明的医疗保健 智能医疗保健
- 卫生事物的互联网 (IoHT)
背景情况:
- 在智能医疗保健中,无处不在的传感能够实现智能心脏声音听觉,但由于智能设备上的敏感数据,它引发了用户隐私问题.
- 联合学习 (FL) 通过在健康事物互联网 (IoHT) 中无需共享数据的去中心化学习来解决隐私问题.
- 传统的FL缺乏模型异质性和客户端个性化,因为它需要在客户端和服务器上统一的架构模型.
研究的目的:
- 为医疗机构的个性化客户模型提出Fed-MStacking,一种使用堆叠合体学习的异质FL框架.
- 解决本地客户端之间不一致的数据标签问题,每个客户端可能只有一个案例类型,无法共享数据.
- 通过汇总缺少的类信息和通过meta-learner构建FL培训的元数据来训练一个全球多类分类器.
主要方法:
- 开发了Fed-MStacking,这是一个异质的FL框架,包含堆叠集体学习.
- 实现了一个meta-learner来汇总缺乏的类信息从客户端与不一致的标签,创建FL的元数据.
- 使用随机森林 (RF),前神经网络 (FNN) 和卷积神经网络 (CNN) 作为多机构心声数据库的基础分类器.
主要成果:
- 在多机构心声分析中,Fed-MSstacking与同质堆叠相比表现优越.
- 该框架成功地支持客户构建个性化,异构的模型.
- 缺乏类信息的聚合使全球多类分类器的培训成为可能,尽管数据不一致.
结论:
- 在智能医疗保健应用程序 (如心声分析) 中,Fed-MStacking为保护隐私,个性化和异质联合学习提供了有效的解决方案.
- 拟议的方法通过利用堆叠集团学习和解决数据异质性和标签不一致性来提高模型性能.
- 该框架促进了FL在卫生事物互联网 (IoHT) 中的应用,用于多机构医疗数据分析.
相关概念视频
Heart Sounds
1.9K
Heart sounds are generated by the turbulence in blood flow due to the closing of heart valves. These sounds are best perceived slightly away from the valves, where the blood flow disseminates the sound.
Auscultation is the process of listening to these internal body sounds using a stethoscope. The heart produces four types of sounds, but only two—S1 and S2—can usually be heard with a stethoscope.
S1, also known as the "lub" sound, is caused by the closure of atrioventricular (A-V)...
Auscultation is the process of listening to these internal body sounds using a stethoscope. The heart produces four types of sounds, but only two—S1 and S2—can usually be heard with a stethoscope.
S1, also known as the "lub" sound, is caused by the closure of atrioventricular (A-V)...
1.9K
Cardiovascular System Abnormal Findings II: Auscultation
124
Auscultation, an essential part of a heart examination, is done using a stethoscope. It provides crucial information about heart function and possible heart problems. Due to heart problems, abnormal sounds can be heard during systole or diastole. These sounds include S3 and S4 gallops, opening snaps, systolic clicks, and murmurs.
Abnormal Heart Sounds
Gallops:
Abnormal Heart Sounds
Gallops:
124
Assessment of the Cardiovascular System IV: Auscultation
280
Cardiac auscultation is a clinical skill used to assess heart function and detect abnormalities. It involves listening to heart sounds at specific anatomical locations through a stethoscope.
Normal Heart Sounds
S1 (First Heart Sound)-
S1 is made by the closure of the mitral and tricuspid valves (atrioventricular valves), marking the beginning of systole.
S2 (Second Heart Sound)-
S2 is made by the closure of the aortic and pulmonic valves (semilunar valves), marking the end of the systole.
Normal Heart Sounds
S1 (First Heart Sound)-
S1 is made by the closure of the mitral and tricuspid valves (atrioventricular valves), marking the beginning of systole.
S2 (Second Heart Sound)-
S2 is made by the closure of the aortic and pulmonic valves (semilunar valves), marking the end of the systole.
280


