Knowledge-Driven Medicine A Machine Learning Approach Chapman & Hall/CRC Data Mining and Knowledge Discovery Series

Focusing on several recent, novel machine learning automatic algorithms, this reference provides the first source on the use of machine learning methods in designing computer-aided diagnosis (CAD) systems. It proposes a framework for CAD problems, presents the technical issues involved when building classifiers, and provides the appropriate machine learning techniques to address these problems. The book also includes results and concrete examples for different diseases and imaging modalities. It provides a useful tool for researchers and students in the biomedical sciences and machine learning as well as for advanced courses on CAD and/or medical informatics.

Introduction

STATE-OF-THE-ART CAD SYSTEMS
General Paradigm of CAD System Design
Lung CAD
Colon CAD
Pulmonary Embolism (PE) CAD
Breast CAD
Automatic heart Wall Motion Abnormality Detection
CAD for Alzheimer’s Disease

TECHNICAL CHALLENGES AND ML-BASED SOLUTIONS
Technical Background
Notation
Formal mathematical definition of the classification problem
Brief intro to classification methods: probabilistic inference, large margin classifiers, and kernel methods

ML Challenges and Novel Solutions
Large number of irrelevant features
Unbalanced data sets
iid assumption is often violated in CAD-related problems
Computational testing efficiency

FUTURE DIRECTIONS AND APPLICATIONS OF CAD SYSTEMS