In this one-semester course, we will explore a variety of advanced machine learning methods for mining clinical data. Much of this exploration will be a hands-on experience, using time in class to expose the principles of each method.
Through a series of lectures and case studies students will gain an understanding of challenges in using machine learning in health care and its application in medicine. The course will cover key use cases such as clinical decision support, personalized medicine and electronic phenotyping.
During the course we will be using R. R is a state of the art environment for statistical computing, which includes a variety of packages for machine learning. Learning objectives for the course: 1.Demonstrate familiarity with the literature on advanced data mining methods 2.Present and discuss the application of methods for mining biomedical data 3.Perform analyses of biomedical data using advanced data mining methods and tools
Prerequisiti
Basic Knowledge of Machine Learning methods. R programming.
Metodi didattici
The course is structured with a series of lectures and lboratories, during which the instructors show the application of the presented methodologies presented to real case studies, using the R.
Verifica Apprendimento
Final Project and Oral Exam The final project comprises two deliverables- a paper and an in-class presentation. The goal of the project is to demonstrate competency in identifying a public-use biomedical or public health dataset, posing a general research question (NOT a hypothesis), proposing and defending an analytic approach to mining the data in order to address the research question, applying the method, evaluating the results, and proposing new directions for further investigation. The project can be done by yourself, or working with another student or in a small team. Each student (or team) will present the final project on the last days of class. Papers will be submitted to the instructors at the end of the course.
The Oral Exam will assess the knowldge of the Methods studied during the course.
Testi
Slides, recorded lectures, and references available on the Kiro page of the course
Contenuti
Clinical Data (Lessons 1 and 2): Data generated by health care systems. Type of clinical data, EHR systems, Taxonomies, common biases in clinical data. Clinical data sharing. Observational Health Data Sciences and Informatics and Federated learning. Precision Medicine. Omics data and unstructured data for precision medicine. How to embed patient generated data in clinical studies.
Statistical Methods: How to handle missing data in clinical datasets via Multiple Imputation by Chained Equations (Lesson 3) + LAB 1 Representing time in clinical data: Mixed Effect Models, Sequential pattern mining, Latent Class Mixed Models (Lesson 4). + LAB 2 Survival analysis, Kaplan–Meier estimator and Cox Regression (Lesson 5) + LAB 3
Use Cases: Electronic phenotyping (Lesson 6) + LAB 4 Clinical studies design (Lesson 7) Clinical Decision Support Systems (Lesson 8)
