ML4H: Machine learning for healthcare

Background:

From self-driving cars to face recognition, machine learning algorithms are being widely applied to enhance human endeavors. Machine learning is so pervasive today that we probably use it dozens of times a day without even knowing it. Many researchers also think it is the best way to make progress towards human-level artificial intelligence. In this day and age, it is therefore critical that we expose students and trainees to this exciting area and educate them in this field. An introductory course such as this one is positioned to train the next generation of medical scientists and practitioners to face data-driven challenges in the coming decades.

The main feature of this course is that students will learn by doing! We will bring machine learning to life by showing fascinating use cases in healthcare and tackling interesting real-world problems like disease risk assessment and predictive modeling. When students complete this introductory course, they will be in a position to analyze several types of medical datasets using machine learning techniques.

Who can take this course?

This course is intended for those who are beginners in machine learning and are interested in applying its concepts to healthcare, including healthcare professionals, students, researchers and data enthusiasts seeking to gain literacy in medical data science.

Pre-requisites:

Basic familiarity with Microsoft Excel or similar analysis software. Prior experience in scientific programing languages (Matlab, Python, etc.) is a plus but not mandatory.

Learning objectives:

• The objective of this course is to teach effective machine learning techniques so that students gain practice implementing them and getting them to work for themselves. Importantly, students will gain the practical know-how needed to quickly apply these techniques to problems in healthcare.

• Students will also get exposure to helpful tools, such as pre-written algorithms, codes and libraries, and to answer interesting questions.

• This course is aimed at machine learning literacy rather than proficiency. Therefore, our intent is to NOT expose students with tedious and rigorous mathematical definitions or equations, but make them comfortable with the tools so that they can tackle a new data challenge without getting overwhelmed with the jargon.

• Students will also learn basic programming skills during hands-on tutorial sessions.

📚 Overview of topics:

Module 1: Introduction to ML in healthcare

- Overview of AI applications in healthcare

- Current trends

Module 2: Data in healthcare

- What’s in the data? Understanding different types of data:
	* Tabular data
	* Medical images
	* Text

- Cohort creation, sampling, risk stratification

- Using Python libraries (pandas, NumPy) for data manipulation

- Data cleaning and preparation techniques specific to healthcare datasets

- Data visualization techniques

Module 3: Fundamental algorithms

- K-means clustering

- Linear Regression

- Classification
	* Logistic Regression
	* Support Vector Machines
	* Decision Trees
	* Random Forest

- Principal component analysis

Module 4: Model development and validation

- Regression analysis 

- Binary classification

- Multiclass and multilabel classification

- Validation techniques: K-fold cross-validation, Confusion Matrix

- Performance metrics: Accuracy, Recall, Sensitivity, F1-Score, AUC ROC, AUC PR

Module 5: Deep learning

- Introduction to neural networks

- Multilayer perceptron

- Generative frameworks

Module 6: Good practices

🙌 Acknowledgments

-- ML4H is maintained by members of the Kolachalama Laboratory.

-- Core contributors and maintainers: Sahana Kowshik, Subhrangshu Bit, Divya Appapogu