HOME: Hands-On Machine Learning

ML Course for Medical Professionals

Ghulam Rasool

Departments of Machine Learning and Neuro-Oncology, Moffitt Cancer Center & Research Institute,

Electrical Engineering Department and Morsani College of Medicine, University of South Florida.

 

______________________________________________________________________________________________________

Motivation

 

This course offers Medical Professionals a practical introduction to Artificial Intelligence and Machine Learning. Through lectures, hands-on labs, and real-world projects, the course attendees will gain valuable skills in harnessing AI towards improving healthcare. The course starts with a grounding in core concepts and methods like regression, classification, and neural networks. With guidance and examples tailored to medicine, techniques for processing medical data and training models to analyze images, text notes, genetic data, and more will be covered. An emphasis is placed on rigorously testing models to ensure they are fair, accurate, and safe for patients. Exciting lectures on state-of-the-art Foundation Models explore how AI breakthroughs like ChatGPT can be applied to tasks like providing patient education or extracting insights from the medical literature. Throughout the curriculum, best practices for developing and deploying reliable, transparent, and ethical medical AI systems are underscored. Upon completing this course, the attendees will have the know-how to implement AI solutions that enhance diagnoses, personalize treatments, streamline workflows, and ultimately get the right care to patients.

 

______________________________________________________________________________________________________

Core Take Aways

Here are some of the core concepts and learnings that Medical Professionals would take away from this proposed course on Machine Learning and AI for healthcare applications:

 

  • Gain practical skills in Python programming, data wrangling, and implementing machine learning pipelines for medical data.

  • Understand how to apply regression, classification, neural networks, CNNs, RNNs, and other core ML techniques to medical use cases.

  • Learn best practices for training, evaluating, and testing models rigorously, focusing on performance metrics relevant to medical practice.

  • Appreciate the expanding capabilities and limitations of state-of-the-art foundation models and large language models.

  • Develop expertise in applying AI to tasks like analyzing medical images and scans, extracting information from clinical text, conversational agents, prediction models, and more.

  • Gain hands-on experience by undertaking an end-to-end ML project from data processing to model deployment.

  • Appreciate the importance of developing interpretable models and integrating AI in clinical workflows responsibly.

  • Become fluent in concepts of data ethics, model bias, regulatory requirements, and other considerations when applying AI in medical settings.

  • Build foundational skills to participate in multidisciplinary teams advancing research and applied AI solutions in healthcare.

  • The core emphasis would be gaining both conceptual and practical AI literacy to responsibly translate machine learning into improved patient outcomes.

 

______________________________________________________________________________________________________

Syllabus

The course comprises of ten basic lectures + labs that cover hands-on ML for Medical Professionals:

Lecture 1: Introduction to Machine Learning for Healthcare

  • Overview of AI/ML applications in medicine
  • Basic concepts like training data, model evaluation, etc.

Lab 1: Intro to Python and ML Libraries

  • Anaconda, Jupyter, Numpy, Pandas, Scikit-learn

Lecture 2: Data Processing and Visualization

  • Working with medical datasets
  • Data cleaning, preprocessing, visualization
 Lab 2: Data Wrangling Medical Datasets
  • Practice data manipulation and visualization

Lecture 3: Linear Regression and Classification

  • Regression vs. classification tasks
  • Models like linear regression, logistic regression
 Lab 3: Building Regression/Classification Models
  • Train and evaluate models on sample datasets

Lecture 4: Trees, Random Forests, Boosting

  • Decision trees and ensemble methods
  • Using tree-based methods with structured data
  Lab 4: Tree-Based Models with Medical Data
  • Experiments with decision trees, random forests

Lecture 5: Neural Networks and Deep Learning

  • Multilayer perceptrons, CNNs, RNNs
  • Applications to images, text, time series, etc.
  Lab 5: Basic Neural Networks with Keras
  • Build and train a simple deep neural network

Lecture 6: Convolutional Neural Networks for Imaging

  • Architecture, training, evaluating CNNs
  • Working with medical images

Lab 6: CNNs for Medical Imaging Data

  • Train a CNN to analyze sample radiology scans

Lecture 7: Recurrent Neural Networks for Sequences

  • RNN architectures like LSTMs and GRUs
  • Sequential data such as patient histories
 Lab 7: RNNs for Clinical Time Series Data
  • Build RNNs to process sample EHR data

Lecture 8: Natural Language Processing

  • Tools for working with clinical notes
  • Topic modeling, information extraction, etc.

Lab 8: NLP for Medical Text

  • Apply techniques to sample notes

Lecture 9: Model Evaluation and Testing

  • Validation, cross-validation, manipulation testing
  • Considerations for medical AI systems
 Lab 9: Evaluating and Testing Medical ML Models
  • Perform validation analyses on sample models
  • Lecture 10: Ethics and Clinical Implementation
    • Principles of ethical medical AI
    • Deploying models to improve care and outcomes
 Lab 10: Final Project
  • End-to-end ML pipeline on a medical use case

 

 

The course also has five additional lectures focusing on Generative AI, Foundation Models, and Large Language Models for medical applications. These lectures will provide a useful introduction to Foundation Models and Large Language Models for healthcare. They cover the core concepts and architectures as well as practical applications and important considerations when deploying these AI systems in clinical settings. The lectures could be paired with labs focused on implementing techniques.

 

Lecture 11: Introduction to Foundation Models

  • What are foundation models?
  • Pre-trained models like BERT and GPT-3
  • Benefits: transfer learning, multi-task learning

Lecture 12: Large Language Models for Biomedical Text Mining

  • LLM architectures (transformer, attention)
  • Pre-trained models like BioBERT, ClinicalBERT
  • Fine-tuning for tasks like entity extraction

Lecture 13: Large Language Models for Medical Conversational Agents

  • Challenges in medical conversation AI
  • LLM-based chatbots for patient education, screening
  • Evaluation of usability, quality, safety

Lecture 14: Foundation Models for Multimodal Healthcare Tasks

  • Integrating images, text, multiple data types
  • Models like CLIP and PathologyCLIP
  • Applications in diagnosis, retrieval, generation

Lecture 15: Considerations in Deploying Foundation Models

  • Challenges with bias, fairness, interpretability
  • Strategies to improve reliability and accountability
  • Importance of ongoing monitoring and evaluation

______________________________________________________________________________________________________

Citation

@misc{home-website,
title = {{HOME: Hands-On Machine Learning Course for Medical Professionals}},
year = {2023},
author = {{Ghulam Rasool}},
note = {Available at: \url{https://lab.moffitt.org/rasool/home/}}
}