Work-in-Progress: Development of an Introductory Machine Learning Course in Biomedical Engineering

Patjanaporn Chalacheva

Download Paper | Permalink

Conference: 2022 ASEE Annual Conference & Exposition
Location: Minneapolis, MN
Publication Date: August 23, 2022
Start Date: June 26, 2022
End Date: June 29, 2022
Conference Session: Biomedical Engineering Division Poster Session
Page Count: 6
DOI: 10.18260/1-2--41468
Permanent URL: https://peer.asee.org/41468
Download Count: 221

Request a correction

Paper Authors

author page

Patjanaporn Chalacheva Carnegie Mellon University

Download Paper | Permalink

Abstract

There has been no shortage of discussion about artificial intelligence and machine learning (ML) in the past decade. ML in healthcare is becoming more widely used and has potential impact on disease prevention and diagnosis. With growing interest in artificial intelligence and machine learning among biomedical engineering students, this work outlines the development of an introductory machine learning course for biomedical engineering (BME) graduate students.

The course introduces high-level concepts behind ML algorithms and teaches students which ML algorithms are best suited to different kinds of biomedical-related problems. A paramount goal is to provide students an appreciation of knowing the “why” and not just the “how” in biomedical data analytics. In general, ML emphasizes the attainment of high accuracy in prediction. This is different in the biomedical field as a key part of the investigation is also to gain insight into or identify the important factors that can explain the underlying processes being studied. Students also learn the importance of pre-screening the data since “garbage in leads to garbage out”, how to choose the most suitable ML algorithm for certain application and know the pitfalls of different methods. While this course is designed to provide BME students their first exposure to ML, it can also serve as the gateway to subsequent specialty ML courses, such as deep learning and probabilistic graphical models, or courses that utilize ML in specific application areas, such as computer vision and bioinformatics.

The target group of this introductory course is primarily incoming graduate students and advanced undergraduates in BME or related disciplines including life science. The broad interdisciplinary background of BME students is the main factor that sets this course apart from machine learning courses traditionally taught in other engineering and computer science programs. A significant proportion of incoming BME Masters students intend to use our BME program as a bridge to switch or expand their base of expertise from biological sciences to engineering. Having a pool of students of non-homogeneous background is thus an additional major challenge in the development and delivery of this course.

The class begins with the basics of probability, statistics, and programming. The rest of the semester covers various supervised and unsupervised learning algorithms with examples drawn from biomedical and life science applications. In addition, students are exposed to biomedical data such as measured physiological signals and medical images where features (certain characteristics of the signal or image) are extracted, and these features are then fed into the ML model. They also see typical issues with biomedical data such as imbalanced datasets in rare diseases or datasets with many missing values. Assessment in this course includes homework (conceptual and coding problems), a midterm exam (conceptual problems) and a group project. The project provides students with an opportunity to get hands-on experience of how one would approach a real-world biomedical problem. Students apply their machine learning knowledge and skills to tackle a biomedical related problem of choice. Assessment mapping, which maps questions on homework, exam and project to the course objectives, is used to evaluate students’ ability in terms of dataset preparation, algorithm selection, algorithm implementation, interpretation of results and model evaluation.

Having offered this course twice, we now have a better understanding of how diverse the student pool is in terms of their previous levels of preparation. Adjustments have been made particularly to coding exercises. Providing optional resources ranging from skeleton codes to short snippets to short hints allows students to choose the level of guidance they need. This work will discuss how the introduction to machine learning for biomedical engineers course was developed, what challenges were encountered and dealt with, and what improvements can be made going forward.

Citation
Format

Chalacheva, P. (2022, August), Work-in-Progress: Development of an Introductory Machine Learning Course in Biomedical Engineering Paper presented at 2022 ASEE Annual Conference & Exposition, Minneapolis, MN. 10.18260/1-2--41468

TY - CPAPER
AB - There has been no shortage of discussion about artificial intelligence and machine learning (ML) in the past decade. ML in healthcare is becoming more widely used and has potential impact on disease prevention and diagnosis. With growing interest in artificial intelligence and machine learning among biomedical engineering students, this work outlines the development of an introductory machine learning course for biomedical engineering (BME) graduate students.

The course introduces high-level concepts behind ML algorithms and teaches students which ML algorithms are best suited to different kinds of biomedical-related problems. A paramount goal is to provide students an appreciation of knowing the “why” and not just the “how” in biomedical data analytics. In general, ML emphasizes the attainment of high accuracy in prediction. This is different in the biomedical field as a key part of the investigation is also to gain insight into or identify the important factors that can explain the underlying processes being studied. Students also learn the importance of pre-screening the data since “garbage in leads to garbage out”, how to choose the most suitable ML algorithm for certain application and know the pitfalls of different methods. While this course is designed to provide BME students their first exposure to ML, it can also serve as the gateway to subsequent specialty ML courses, such as deep learning and probabilistic graphical models, or courses that utilize ML in specific application areas, such as computer vision and bioinformatics.

The target group of this introductory course is primarily incoming graduate students and advanced undergraduates in BME or related disciplines including life science. The broad interdisciplinary background of BME students is the main factor that sets this course apart from machine learning courses traditionally taught in other engineering and computer science programs. A significant proportion of incoming BME Masters students intend to use our BME program as a bridge to switch or expand their base of expertise from biological sciences to engineering. Having a pool of students of non-homogeneous background is thus an additional major challenge in the development and delivery of this course.

The class begins with the basics of probability, statistics, and programming. The rest of the semester covers various supervised and unsupervised learning algorithms with examples drawn from biomedical and life science applications. In addition, students are exposed to biomedical data such as measured physiological signals and medical images where features (certain characteristics of the signal or image) are extracted, and these features are then fed into the ML model. They also see typical issues with biomedical data such as imbalanced datasets in rare diseases or datasets with many missing values. Assessment in this course includes homework (conceptual and coding problems), a midterm exam (conceptual problems) and a group project. The project provides students with an opportunity to get hands-on experience of how one would approach a real-world biomedical problem. Students apply their machine learning knowledge and skills to tackle a biomedical related problem of choice. Assessment mapping, which maps questions on homework, exam and project to the course objectives, is used to evaluate students’ ability in terms of dataset preparation, algorithm selection, algorithm implementation, interpretation of results and model evaluation.

Having offered this course twice, we now have a better understanding of how diverse the student pool is in terms of their previous levels of preparation. Adjustments have been made particularly to coding exercises. Providing optional resources ranging from skeleton codes to short snippets to short hints allows students to choose the level of guidance they need. This work will discuss how the introduction to machine learning for biomedical engineers course was developed, what challenges were encountered and dealt with, and what improvements can be made going forward.
AU - Patjanaporn Chalacheva
CY - Minneapolis, MN
DA - 2022/08/23
PB - ASEE Conferences
TI - Work-in-Progress: Development of an Introductory Machine Learning Course in Biomedical Engineering
UR - https://peer.asee.org/41468
DO - 10.18260/1-2--41468
ER -