Forghani R, ed. Mukherji SK, consulting ed. Neuroimaging Clinics of North America: Machine Learning and Other Artificial Intelligence Applications. Elsevier; 2020;30(4):393–530; $397.00
Machine Learning and Other Artificial Intelligence Applications is a thoughtfully arranged collection of review articles offering an overview of not only artificial intelligence (AI), but also AI in neuroradiology. The perspective ranges from a broad historical view to specifics of the most pertinent and modern techniques for the incorporation of AI into radiology. It has an eye on the past and on the future; going through the collection offers a great way to learn the basics of AI and its context in radiology.
It is easy to imagine that many of the people with an interest in AI likely have an interest in the future but not necessarily the past. Fortunately, the authors of the articles selected for historical context have done a good job of not only presenting the pertinent history of AI, but also doing so in a way that points towards its potential and the excitement surrounding that. For example, the first article, “Brief History of Artificial Intelligence,” covers the origins of AI, the Turing test, and other features of the path that has brought us to where we are today, while also giving insight into how expectations play a role in the growth and progression of the AI field. With this, the stage is set for the reader to learn not only more about the basics of AI, but also how it may succeed or falter. Additionally, there are succinct and effective explanations of deep learning and machine learning as well as the role of convolutional neural networks. This allows the reader to prepare for more detailed and technical articles which come later in the volume.
Most of the technically oriented articles focus on a particular application of AI, its methods, and its performance. In doing so, the authors delve into deep details of convolutional neural networks. They offer explanations of how these are designed and function, with articles offering examples and illustrations of the different ways neural networks may be arranged and operated. Many of the lessons found in this volume reach a level of detail appropriate for physicians to follow but also dig deep enough to lead those who are interested in a path of becoming a functional contributor to AI development as it becomes further entwined with radiology. Towards the other end of the spectrum, there are some articles or portions of articles that come just shy of coding. These can be fascinating but also abstract. The proportion of the book dedicated to such detail appears suitably balanced to allow those who are truly technically inclined to learn more while not overwhelming most readers.
A praiseworthy aspect of the book are the illustrations found in the articles. The individual article authors succeeded in offering figures that help clarify what is innately an esoteric topic. They integrate medical imaging with explanatory models for neural networks in a way that is grounding to the radiologist who chooses to flip through the pages. The print quality of some of the images is sufficient to make up for some of the smaller figures. There is adequate contrast for appreciation of the details in the medical images. Of note, perusing the figures in the book may not be a sufficient review for those who wish to revisit the concepts at a later time. This may be a byproduct of the subject matter or speak to the need for more illustrations.
The editor also chose some articles with a broader view of AI applications, allowing for readers to be introduced to different areas of medical AI which may have escaped them while reading the barrage of AI-related headlines in industry emails, newsletters, and social media. The article “Diverse Applications of Artificial Intelligence in Neuroradiology” offers a great list of applications for AI that physicians may not be aware are being explored. These include triaging cases, improvements in education, disease quantification, and protocoling.
There is no doubt that the more forward-looking aspects of the book will not age well, as is inherent in any review of a technology. But currently, it compares favorably to other outlets for learning about AI. An intrinsic strength to its format is its function as an introduction to many of the important voices in medical AI. Readers will not only become familiar with the subject material, but also with the contributors to the field and their perspectives.
One aspect that appears to be lacking is a section of the book that would unify many of the varied concepts addressed in individual articles. An introductory chapter beyond the preface and foreword would help lend a unifying voice to the collection, which otherwise feels unguided. Additionally, a glossary would be a useful resource while reading the volume and would also add lasting value to its real estate on the reader’s bookshelf.
Overall, Machine Learning and Other Artificial Intelligence Applications is a well-curated compilation of articles for those looking to become knowledgeable in AI. It is a well-thought-out shortcut to understanding AI, offering a quicker means to competence when compared with unguided web searches and hunting for articles in periodicals. The illustrations are thoughtful and the authors knowledgeable. It is an ideal book for early-career radiologists who realize that they will need to master AI to succeed, as well as a book for experienced radiologists who want to stay ahead of the curve.