BMI 6334 - Deep Learning in Biomedical Informatics

3 semester credit hours
Lecture contact hours: 2; Lab contact hours: 3
Web-based and classroom instruction
Prerequisite: BMI 5007 and BMI 5353 or Instructor Consent

Course Description

Deep learning and artificial intelligence have disrupted multiple industries including healthcare. This class offers students exposure to basic concepts of and practical skills for deep learning and its applications in selected problems in biomedical informatics. Students will study the foundations of deep learning, understand how to build neural networks, and conduct successful machine learning analyses. Deep learning architectures such as convolutional neural networks, recurrent neural networks, and autoencoders will be explored, along with concepts such as embeddings, dropout, and batch normalization. Case studies from biomedical informatics, including biomedical and clinical natural language processing, medical imaging, electronic health records, and genomics data will be utilized. Students will use the Python language and the state-of-the-art deep learning frameworks to implement deep learning models to solve real world problems. Experience with Python programming and basic knowledge of linear algebra is required.

Learning Objectives

Upon successful completion of the course, students will

• Discuss the basic concepts of artificial neural networks.
• Determine the applicability of deep learning approaches for real biomedical informatics problems.
• Develop hands-on programming skills using a deep learning framework, e.g., TensorFlow/Keras/PyTorch.
• Build, debug, and visualize a convolutional neural network for medical image analysis and sequential models for biomedical NLP analysis, as well as a predictive model using structured electronic health record and genomic data.