Reviewer: Hui Lin, PhD (University of Pennsylvania School of Medicine)
Description: This multiauthored book covers basic and advanced techniques in the field of outcome modeling in radiation therapy, providing a comprehensive introduction to the current state of data resources, statistical and numerical aspects of modeling approaches, and clinical applications and challenges of outcome modeling. It is well edited and easy to read, and it provides useful references and code examples for follow-up and hands-on practice.
Purpose: The field of radiation therapy has seen emerging growth in various formats of clinical, volumetric imaging, dosimetry, and genomic data. Parallel developments in advanced statistical and machine learning methodologies, radiomics, and big data processing techniques enable the fusion of data and model to advance the outcome studies of radiotherapy treatment. This book fills a niche in this field, as it provides a comprehensive overview of data resources that can be used for outcome model building, introduces statistical and numerical modeling methods, and offers in-depth discussions on active research topics such as machine learning and radiomics. The book communicates the topics well, providing reasonable detail about theoretical background to help readers understand the analytical, statistical, and numerical modeling methods related to outcome analysis and prediction, while maintaining the characteristics of a practical handbook by serving code examples ranging from data exploration to model building and evaluation.
Audience: This book is recommended primarily for researchers in outcome modeling in radiation therapy who want a comprehensive overview of resources on outcome modeling and be informed about the up-to-date developments in key areas of outcome modeling research. In addition, the theoretical details of modeling and the practical code examples provide an excellent entry point for trainees and practitioners to explore the outcome modeling studies. Many chapters of the book can be compiled into a textbook supplement for classes in advanced statistics and modeling topics for medical physics graduate students and residents. All the contributors are respected experts in the outcome modeling field. The author is an internationally recognized expert in bioinformatics, machine learning, and outcome modeling with a broad spectrum of clinical applications.
Features: The first of the book's four sections introduces the general workflow of outcome modeling and different types of data that can be incorporated into model design and building, such as clinical data, imaging data, dosimetry data, and biological data. Chapters in this section discuss the extraction, use, and impact of each type of data and provide a solid foundation for understanding the essence of outcome modeling along with a comprehensive review of related work to date. The next two sections go into depth about specific methods of modeling in radiotherapy outcome analysis. The top-down modeling section presents a thorough description of data analysis methods including mechanistic, statistical regression, and machine learning models. The approaches to model selection, evaluation, and feature selection are robustly illustrated, followed by accessible reviews summarized by the authors and code examples. The level of detail in these chapters is sufficient to provide a fundamental understanding of the essential elements of top-down modeling and important concepts of statistics. The bottom-up modeling section covers the stochastic and nonstochastic methods for modeling the multiscale biological response in radiation therapy, chemotherapy, and immunotherapy. These chapters also include useful introductions to software where the featured models can be implemented. The last section discusses the extent of existing and emerging applications of outcome modeling ranging from treatment planning, adaptive therapy, and particle therapy to clinical trials. The power of outcome modeling as well as its relevance to the radiation therapy clinical community is demonstrated. The code examples embedded after most of the modeling theories is a great strength of the book, providing not only additional clarifications of the methods, but also hands-on exercises for readers to learn the featured models. A shortcoming of the book is the omission of an introduction to advanced artificial intelligence algorithms for outcome modeling. The use of artificial intelligence methods, including deep learning and reinforcement learning methods, has emerged as a fast-growing trend for outcome analysis and prediction, hence it is necessary to cover and emphasize these methods for educational purposes. Another minor pitfall is that several figures are not plotted in high-resolution vector formats, making them blurry.
Assessment: This book is easy to read while providing comprehensive theoretical detail on outcome modeling relevant to the radiation therapy community. The code examples provided in almost every chapter are invaluable in clarifying the concepts and are excellent starting points for students and practitioners to get some practical experience. Each chapter is well written by recognized experts with excellent illustrations and topical reviews of different aspects of outcome modeling. I would recommend this book to anyone wishing to get a thorough understanding of the fundamental areas in radiotherapy outcome modeling.