Machine learning-based brain aging morphological signatures of chronic musculoskeletal pain

Project Summary/Abstract. Chronic musculoskeletal (MSK) pain is experienced by millions in the United States (US), with highest prevalence in older adults, leading to disability. An approach to attain improvement in diagnosis and treatment that has gained momentum over the last decade is to increase understanding of the neural underpinnings of

chronic pain by looking at the associated brain changes. Biomarkers of healthy brain aging, derived from machine learning (ML) models that map MRI patterns to chronological age, offer a look at these pathological changes since deviations from healthy brain aging indicate potential underlying pathologies. The Predicted brain Age

Difference (PAD; predicted brain age minus chronological age), proposed as a biomarker of disease, was found to be higher with the presence, and positively correlated with the severity, of MSK pain. However, because it is

based on a ‘global’ brain age measure, the PAD is limited to signal a ‘poorer health’ state without specifying the type of underlying pathology. This project proposes to develop novel spatially distributed brain age measures (brain age maps) able to capture the brain atrophy signature of different MSK conditions like chronic back pain,

osteoarthritis and neck or shoulder pain. These brain age maps will be obtained from T1-weighted MRIs via very innovative convolutional neural network (CNN) architectures that fuse local and global mechanisms of contextual information in the image using the so-called “transformers”. The project then proposes to develop biomarkers

specific to these MSK types informed by the brain age maps via CNNs and the so-called “vision transformers”, a cutting-edge methodology ideal for image classification. By accomplishing these goals, this project will reveal useful information about distinct neurobiological mechanisms of different MSK types and their determinants (e.g.,

aberrant sensory testing or resting functional networks), and how brain age is associated to the multidimensional experience of pain. This could be particularly useful to understand the causes of the high MSK prevalence in older adults. For example, the highest MSK prevalence in older adults might be the consequence of a more

‘natural’ accelerated brain aging, in contrast to more ‘insult-like’ causes in younger adults. Thus, the project also aims to investigate possible age-related differences in PAD maps of MSK. Finally, we evaluate the brain age maps’ ability prognosticate chronic pain chronification and pain-related functional decline. This proposal is

powered by the tens of thousands of participants with MRI and pain data in the UK Biobank and leverages the University of Florida’s Artificial Intelligence (AI) Initiative, endowed with one of the most powerful high performance computing infrastructures across universities in the US. With this significant study, the applicant

pursues an independent career as an expert in ML/AI methodologies to be used to identify novel pain and brain aging biomarkers. He will also boost his understanding of experimental pain and of the neurobiological mechanisms of pain and aging. He will leverage strong institutional support, stellar mentorship in pain, aging and

AI, a thriving collaborative AI network, and powerful computing resources.