Aging Health Behind an Image: Quantifying Sarcopenia and Associated Risk Factors from Advanced CT Analysis and Machine Learning Technologies

Sarcopenia, the progressive degeneration of aging muscle, is identified as an independent risk factor for significant morbidity, disability, and mortality in elderly individuals. In this paper Artificial Intelligence technologies, in particular Machine Learning (ML) supervised algorithms, are adopted to predict physiological parameter starting from muscle, fat and connective tissue distribution values of a mid-thigh Computer Tomography (CT) images. We developed and validated a novel method for soft tissue radiodensitometric distribution profiling, which is entitled nonlinear trimodal regression analysis (NTRA) method for soft tissue CT profiling. The work shows a comparative analysis using the NTRA method and standard soft tissue CT analysis modalities which was implemented on parameters assemblies from the 3,157 patients AGES-Reykjavik dataset. Furthermore, ML approach is used to find connections between amplitude, location, width and skewness in fat, muscle, and connective tissue and link these data to biomechanical measurements, Body Mass Index (BMI) and Cholesterol. The results highlight the specificities of each muscle quality metric to Lower Extremity functions and sarcopenic comorbidities. ML approach shows good predictive values for BMI having as most significant features connective and fat amplitude. Standardizing a quantitative methodology for myological assessment in this regard would allow for the generalizability of sarcopenia research to the indication of compensatory targets for clinical intervention.