There has been growth in recent years in the application of artificial intelligence (AI) and machine learning (ML) algorithms in biomedicine. This growth is most pronounced in areas related to radiological applications and medical physics, including the publication of special issues that graced the pages of Medical Physics. This growth has inadvertently led to inconsistent reporting of AI/ML findings in the literature, confusing the interpretation of their results and eroding trust in their potential impact. 

Therefore, several leading societies and peer-reviewed journals have developed recommendations to improve the reporting of AI/ML studies for biomedical applications and to restore trust in their published findings. For instance, Radiology has developed a checklist for AI in medical imaging (CLAIM) that followed the format of EQUATOR Network guidelines and also incorporated general manuscript review criteria for diagnostic imaging applications. Nature Medicine also published a minimum information checklist for clinical AI modeling (MI-CLAIM).

Medical Physics, as the flagship scientific journal of the American Association of Physics in Medicine (AAPM) and its sister organizations, publishes theoretical and experimental research on applying physics, mathematics, and engineering to solve problems in medicine and human biology. AI/ML applications are currently playing a pivotal role across the entire landscape of our discipline. In keeping with other journals, we introduce a new, required checklist for AI/ML applications in Medical Physics (CLAMP) to ensure rigorous and reproducible research of AI/ML in medical physics. 

Assessing MR Artifacts

As clinical magnetic resonance (MR) imaging becomes more versatile and complex, it is increasingly difficult to develop and maintain a thorough understanding of the physical principles governing the changing technology. This is particularly true for practicing radiologists, whose primary obligation is to interpret clinical images and not necessarily to understand complex equations describing the underlying physics. 

Nevertheless, the physics of MR imaging plays an important role in clinical practice because it determines image quality and suboptimal image quality may hinder accurate diagnosis. This article provides an image-based explanation of the physics underlying common MR imaging artifacts, offering simple solutions for remedying each type of artifact. 

Solutions that have emerged from recent technological advances with which radiologists may not yet be familiar are described in detail. Types of artifacts discussed include those resulting from voluntary and involuntary patient motion, magnetic susceptibility, magnetic field inhomogeneities, gradient nonlinearity, standing waves, aliasing, chemical shift, and signal truncation. With improved awareness and understanding of these artifacts, radiologists will be better able to modify MR imaging protocols to optimize clinical image quality, allowing greater confidence in the diagnosis.

Role in Radiation Oncology

Medical physics has a long tradition of contributing to modeling biological effects in radiation oncology. High-impact examples are quantifying dose-volume effects based on clinical data, being relevant in everyday radiotherapy planning and optimization, and the adaptation and use of fractionation models aimed to translate physical doses into biologically equivalent doses for tumors.

Medical physicists have fundamental physics skills to set up mathematical descriptions of biological or clinical problems, combined with the ability to simplify complex relationships to the greatest extent. In addition, medical physics training in fundamental mathematical, statistical, biological and clinical aspects allows medical physicists to relatively easily interact with the professionals required for successful interdisciplinary teams to tackle modeling problems. Machine learning and AI-based models derived from data can be useful, but an appropriate level of understanding and extensive validation is needed to give sufficient confidence for clinical use.

Along with AI implementation, medical physicists should also act as facilitators of data gathering and data farming, contributing to building and managing advanced data-sharing platforms and within new approaches such as umbrella protocols and basket trials. 

Conclusion

For AI/ML applications in medical physics, a problem statement and rationale for utilizing these algorithms are necessary while highlighting the novelty of the approach. A brief numerical description of how the data are partitioned into subsets for the AI/ML algorithm training, validation, and independent testing of algorithm performance is required. This is to be followed by a summary of the results and statistical metrics that quantify the performance of the AI/ML algorithm.