依托南京医科大学国际课程平台，南京医科大学生物医学工程与信息学院特邀法国雷恩大学居伊·卡罗（Guy Carrault）教授开设国际课程《机器学习：生物医学信号实践》。课程计划于2026年8月26日正式开课，将围绕机器学习在生物医学信号分析中的实践应用展开系统讲授与实操训练。
Relying on the international course platform of Nanjing Medical University, the College of Biomedical Engineering and Informatics of Nanjing Medical University invites Professor Guy Carrault from the University of Rennes, France, to teach the international course Machine Learning: Apply to Biomedical Signal. The course is scheduled to start on August 26, 2026, and will provide systematic lectures and hands-on training on the practical application of machine learning to biomedical signal analysis.
课程描述 Description of Course
本课程主要面向生物医学工程、智能医学工程、生物与医药等相关专业研究生，围绕机器学习在生物医学信号分析中的实践应用展开。课程将系统介绍生物医学信号分类策略、特征提取、模式识别、无监督学习、监督学习、模型性能评价及特征排序等核心内容，并结合室性心律失常预测、早产新生儿脓毒症预测和基于加速度计的日常活动识别等真实案例开展教学。课程采用理论讲授与实践训练相结合的方式，引导学生掌握从生物医学数据处理、特征分析到机器学习建模与结果解释的完整流程。课程共21学时，其中理论课14学时、实践课7学时，学习完成者可获得2学分。
This course is primarily designed for graduate students majoring in Biomedical Engineering, Intelligent Medical Engineering, Biology and Medicine, and related disciplines. It focuses on the practical application of machine learning to biomedical signal analysis, covering key topics such as biomedical signal classification strategies, feature extraction, pattern recognition, unsupervised learning, supervised learning, model performance evaluation, and feature ranking. The course will incorporate real-world case studies, including ventricular arrhythmia prediction, sepsis prediction in premature newborns, and accelerometer-based daily activity recognition. By combining theoretical lectures with hands-on training, the course guides students through the complete workflow from biomedical data processing and feature analysis to machine learning modeling and result interpretation. The course consists of 21 class hours, including 14 hours of lectures and 7 hours of practical training, and participants who complete the course will be granted 2 academic credits.
教授简介 Professor introduction
居伊·卡罗（Guy Carrault）教授现任法国雷恩大学教授，长期从事生物医学工程、信号处理、检测与估计、基于人工智能的决策系统、心率变异性分析、重症监护监测以及新型医疗设备开发等方向的教学与科研工作。卡罗教授于1987年获法国雷恩大学博士学位，并于1997年获得指导研究资格认证（HDR）。自2000年起，他担任法国雷恩大学教授，具有深厚的生物医学信号处理与机器学习研究基础，以及丰富的国际化教学经验。
卡罗教授曾任LTSI-INSERM 1099联合研究单位副主任、LTSI-INSERM 1099单位SEPIA团队负责人，以及Inserm认证临床研究中心技术创新平台主管。同时，他在教学管理方面具有丰富经验，曾担任雷恩大学电气工程与嵌入式系统系主任、机电一体化职业本科项目负责人、SISEA硕士项目负责人，以及本科电子与嵌入式系统专业方向负责人。
在学术与科研方面，卡罗教授长期活跃于生物医学信号处理、医学人工智能、数字健康与医疗器械创新等领域。他曾担任法国国家大学理事会信号处理学科委员，参与GDR STIC-Santé指导委员会工作，并长期参与博士生培养与国际学术合作。他还担任多个国际期刊审稿人，包括 IEEE Transactions on Biomedical Engineering、Medical & Biological Engineering & Computing、Artificial Intelligence in Medicine、Digital Signal Processing 等。
卡罗教授主持或参与多项法国国家级、欧盟及国际合作科研项目，包括欧洲Digi-NewB项目、PEPR Santé Numérique DIIP-Heart项目、AMI-CMA ARCLIMED项目，以及多项法国国家科研署（ANR）项目。在产业转化方面，他拥有19项已注册专利和16项软件登记，其中多项成果已实现企业转化。近年来，他在生物医学信号处理、心率变异性分析、新生儿监测、睡眠呼吸事件检测、医学人工智能建模等方向持续发表高水平研究成果，相关论文发表于 IEEE Journal of Biomedical and Health Informatics、IEEE Transactions on Biomedical Engineering、Biomedical Signal Processing and Control、International Journal of Medical Informatics 等国际期刊。
Professor Guy Carrault is currently a Professor at the University of Rennes, France. His long-standing teaching and research activities focus on biomedical engineering, signal processing, detection and estimation, AI-based decision systems, heart rate variability analysis, monitoring in intensive care units, and the development of new medical devices. He received his PhD from the University of Rennes in 1987 and obtained the Habilitation to Direct Research in 1997. Since 2000, he has served as a Professor at the University of Rennes, with extensive expertise in biomedical signal processing, machine learning, and international education.
Professor Carrault has held a number of important academic and research positions, including Deputy Director of the LTSI-INSERM 1099 Joint Research Unit, Head of the SEPIA team at LTSI-INSERM 1099, and Scientific Manager of the Inserm-labeled Clinical Investigation Center for Technological Innovation. He also has extensive experience in educational leadership, having served as Head of the Electrical Engineering and Embedded Systems Department, Head of the Professional License Program in Mechatronics, Head of the SISEA Master’s Program, and Head of the Electronic and Embedded Systems pathway at the University of Rennes.
In terms of academic and research contributions, Professor Carrault has been actively involved in biomedical signal processing, medical artificial intelligence, digital health, and medical device innovation. He has served as a member of the French National University Council in the signal processing section, participated in the Steering Committee of GDR STIC-Santé, and has been deeply involved in doctoral training and international academic collaboration. He also serves as a reviewer for many international journals, including IEEE Transactions on Biomedical Engineering, Medical & Biological Engineering & Computing, Artificial Intelligence in Medicine, and Digital Signal Processing.
Professor Carrault has led or participated in numerous French national, European, and international collaborative research projects, including the European Digi-NewB Project, the PEPR Santé Numérique DIIP-Heart Project, the AMI-CMA ARCLIMED Project, and multiple projects funded by the French National Research Agency. In industrial translation, he holds 19 registered patents and 16 software registrations, several of which have been transferred to industry. In recent years, he has continued to publish high-level research in biomedical signal processing, heart rate variability analysis, neonatal monitoring, sleep apnea detection, and medical AI modeling, with publications in journals such as IEEE Journal of Biomedical and Health Informatics, IEEE Transactions on Biomedical Engineering, Biomedical Signal Processing and Control, and International Journal of Medical Informatics.
讲课内容 Course Outline and Assignment
1）机器学习与模式识别基础：课程将首先介绍机器学习平台的整体框架，系统讲解机器学习在生物医学信号处理中的应用逻辑，帮助学生理解从原始信号到分类决策的完整模式识别链路。同时，课程将围绕特征提取展开基础教学，使学生掌握从生理信号中提取有效变量的基本思路与方法。
2）生物医学信号案例教学：课程将结合两个真实生物医学案例开展教学。第一个案例为基于植入式心律转复除颤器时间序列数据的室性心律失常近期预测，帮助学生理解机器学习在心脏事件风险预测中的应用。第二个案例为早产新生儿脓毒症预测，引导学生认识机器学习方法在新生儿重症监测和临床辅助决策中的潜在价值。
3）描述性方法与主成分分析：系统讲解连续域与离散域维纳滤波的原理与实现方法，配套线性预测相关内容教学，通过专项习题课深化学生对AR模型、最优滤波核心知识点的理解与应用能力，夯实最优滤波技术基础。
4）无监督学习方法：深入讲解自适应滤波的核心原理与最小均方（LMS）算法，拓展滤波技术在生物医学信号处理、语音降噪等领域的进阶应用，配套完整的实操教学环节。
5）数据预处理与性能评价：通过基于Matlab软件的实操课程，指导学生完成基于自适应滤波的降噪处理全流程开发，实现课程理论知识的工程化落地，完成课程实践教学目标。
6）监督学习方法：课程将重点讲解监督学习方法，包括K近邻、决策树和随机森林等经典机器学习算法。K近邻部分将介绍K值选择及其对分类性能的影响；决策树部分将讲解训练过程、熵、基尼不纯度、信息增益、预剪枝与后剪枝、过拟合控制及超参数选择；随机森林部分将介绍集成学习思想、模型工作原理及超参数调节方法。配套实践课程将帮助学生掌握这些方法在日常活动识别任务中的具体应用。
7）特征排序与课程实践落地：课程最后将介绍特征排序方法，包括信息增益、Morris分析和ReliefF方法，帮助学生理解不同特征对模型分类性能的贡献。通过完整实践训练，学生将围绕基于加速度计的日常活动识别任务，完成从特征提取、数据处理、模型训练、性能评价到特征解释的完整机器学习应用流程，实现课程理论知识向实际生物医学信号分析任务的落地转化。
1）Fundamentals of Machine Learning and Pattern Recognition: The course will begin with an overview of the machine learning platform and systematically introduce the application logic of machine learning in biomedical signal processing. Students will be guided to understand the complete pattern recognition pipeline from raw signals to classification decisions. The course will also cover the fundamentals of feature extraction, helping students master the basic principles and methods for extracting informative variables from physiological signals.
2）Case Studies in Biomedical Signal Analysis: The course will include two real-world biomedical case studies. The first case focuses on the short-term prediction of ventricular arrhythmia based on time-series data from implantable cardioverter-defibrillators, helping students understand the application of machine learning in cardiac event risk prediction. The second case focuses on the prediction of sepsis in premature newborns, illustrating the potential value of machine learning methods in neonatal intensive care monitoring and clinical decision support.
3）Descriptive Methods and Principal Component Analysis: This module introduces descriptive machine learning methods, with a particular focus on the principles, result interpretation, and application scenarios of Principal Component Analysis (PCA). Through hands-on practical sessions on accelerometer-based daily activity recognition, students will learn how to use PCA for dimensionality reduction, visualization, and feature interpretation, and understand how different variables influence classification and recognition results.
4）Unsupervised Learning Methods: This module systematically introduces unsupervised learning methods, including K-means clustering and hierarchical clustering. Key topics include the determination of the number of clusters, the importance of distance metrics, and hierarchical clustering strategies such as Ward’s method, complete linkage, average linkage, and single linkage. Through practical training, students will apply clustering methods to accelerometer-based activity recognition tasks and complete activity classification, feature description, and result interpretation.
5）Data Preprocessing and Performance Evaluation: This module explains the importance of data scaling, normalization, and standardization in machine learning modeling. It also introduces common performance evaluation metrics for classification models, including confusion matrices, accuracy, recall, ROC curves, AUC, and multi-class classification evaluation. Through this module, students will understand the importance of standardized data processing and scientific model evaluation in practical machine learning applications.
6）Supervised Learning Methods: This module focuses on supervised learning methods, including K-nearest neighbors, decision trees, and random forests. The K-nearest neighbors section will explain the selection of K values and their influence on classification performance. The decision tree section will cover the training process, entropy, Gini impurity, information gain, pre-pruning and post-pruning, overfitting control, and hyperparameter selection. The random forest section will introduce ensemble learning, model principles, and hyperparameter tuning. Practical sessions will help students apply these methods to daily activity recognition tasks.
7）Feature Ranking and Practical Course Implementation: The final module introduces feature ranking methods, including Information Gain, Morris analysis, and ReliefF. These methods will help students understand the contribution of different features to model classification performance. Through complete practical training, students will complete the full machine learning application process for accelerometer-based daily activity recognition, including feature extraction, data processing, model training, performance evaluation, and feature interpretation, thereby translating theoretical knowledge into practical biomedical signal analysis applications.
报名与安排 Registration and Arrangement
如果您想要报名，可通过学院指定官方渠道完成报名。本课程学习免费，主要面向生物医学工程、智能医学工程、生物与医药等相关专业硕士研究生一年级开设，计划招收30名学生。课程授课时间为2026年8月26日、8月27日、8月28日、8月31日及9月1日，授课地点为学海楼D114。如有其他问题，可联系课程协调专员或课程助理咨询相关信息。
If you would like to register, please complete the registration through the official channel designated by the college. This course is free of charge and is mainly open to first-year master’s students majoring in Biomedical Engineering, Intelligent Medical Engineering, Biology and Medicine, and related disciplines. The course plans to enroll 30 students. Classes will be held on August 26, August 27, August 28, August 31, and September 1, 2026, in Room D114, Xuehai Building. For further questions, please contact the course coordinator or course assistant for more information.
联络人
协调专员：向文涛
课程助理：宋子晨（Tel/Wechat: 15190755518）
授课地点：学海楼D114
Contacts
Coordinator: Wentao Xiang
Assistant: Zichen Song（Tel/Wechat: 15190755518）
Location: Room D114, Xuehai Building
发布审核：刘继永、高珍秀
