This course is designed for students who are passionate about innovative technologies and interested in applying data processing principles to health care problems. The course provides a comprehensive overview of practical data processing methods and applications in medicine. Students will acquire a comprehensive set of information and tools that enable them to process medical data, extract meaningful information, select important features, classify categories, and predict outcomes. The course follows the flow of data from quality checks to major preprocessing steps. Machine learning models will be introduced, from simple to complex, comparing architectures and providing justifications for model selection. In addition, recommendations will be provided on how to train and optimize machine learning models and how to evaluate and compare their performance. Case studies and real-world practical examples are included, comprising analysis of structured/tabular data (i.e., electronic medical records) and one-dimensional/multidimensional time series (i.e., physical and physiological signals). In addition, common errors in data processing will be highlighted, as well as incorrect model training and evaluation strategies. Finally, students will be asked to apply the acquired knowledge to develop their own classification system using health-related data of their choice.

Basic knowledge of statistics, linear algebra and calculus

• Introduction - Significance of medical data analysis in advancing healthcare - Understanding different types of medical data: physical/physiological signals and electronic health records - Machine learning applications in health monitoring and decision-making - Requirements for successful medical data analysis: affordability, interpretability, and ethics - Familiarization with open-access medical datasets for research and development • Data preprocessing - Detecting outliers and addressing their impact on analysis - Filtering for noise reduction - Segmentation of medical signals - Data representation and transformation - Feature extraction methods for relevant information retrieval - Feature selection techniques for enhanced model performance - Class balancing methods to address class imbalance • Classification and Regression - Assessing data size and distribution for model selection - Support vector machines and k-nearest neighbors - Artificial neural networks in healthcare - Exploring deep learning methods: convolutional and recurrent neural networks - Promoting generalization capability • Performance evaluation - Validation methods: general vs subject-specific models - Classification metrics: evaluating model performance in healthcare - Regression metrics: assessing predictive accuracy for medical outcomes

Modalità mista

Sviluppo di project work in team - Presentazione orale

P.D.2-2 - Maggio