Advancing Multiple Myeloma Prognostication and Monitoring through AI-Powered Comprehensive Imaging - Summary Multiple myeloma (MM) is a hematologic cancer characterized by abnormal plasma cells in the bone marrow, affecting 0.76% of the U.S. population with over 35,000 new cases annually; despite being historically considered incurable, recent treatment advances now achieve a median progression-free survival of 41 months. Diagnostic imaging has evolved significantly, with Whole Body MRI (WB-MRI) critical for initial staging and monitoring of MM, where WB diffusion-weighted imaging (WB-DWI) excels in lesion detection and treatment response assessment. These imaging techniques are complemented by the Dixon SPINE protocol, which measures high fat fractions (FF) associated with deeper responses in MM patients, and the assessment of sarcopenia and bone marrow density (BMD), which are crucial but underexplored indicators of overall survival and disease progression. Advanced deep learning algorithms are being developed to enhance the segmentation and quantification of these imaging biomarkers, aiming to create predictive, multi-modal AI models to improve MM patient management and outcomes. This project aims to create deep learning models to enhance the management of MM, focusing on automating the quantification of disease burden and developing predictive models for patient outcomes. Aim 1 aims to develop automated segmentation tools using WB-MRI to identify key biomarkers. such as bone lesions, FF, sarcopenia, and BMD, for assessing disease progression. Aim 2 builds on the biomarkers identified in Aim 1, integrating them with clinical, biological, and genetic factors to dynamically assess patient outcomes, including treatment response and overall survival, following the standards of the International Myeloma Working Group (IMWG). This aim involves developing longitudinal AI models that could revolutionize personalized medicine by predicting disease trajectories and treatment responses. Aim 3 involves a prospective study to validate the created tools and predictive models in real clinical settings. In summary, the goal of this project is to develop and validate advanced AI-driven tools for automated disease burden quantification, establishing biomarkers in multiple myeloma, and for creating multi-modal AI models to predict patient outcomes on longitudinal data. The significance of this work lies in its potential to dynamically and quantitatively enhance prognostic assessments and monitoring for multiple myeloma patients, leading to more precise and effective management strategies.
