Sonification and Clustering: a Multi-Sensory Perspective for Precision Medicine

Melodies are sequences of pitches distributed over time. In medicine, clinical variables, measured at regular intervals, can also be treated as quantities distributed over time. Thus, they can be mapped to sounds, creating melodies. These sonified trajectories can provide a multi-sensory perspective on disease progress for a sample of patients. Musical similarity becomes a tool to group patients with similar disease progress. Here, we focus on diabetic kidney disease (DKD) patients, proposing a sonification of their filtering efficiency information (through eGFR). We obtain Hertz for pitches and spectral centroids. Then, we use this information to build clusters of patients. We adopt a statistics method based on curve-parameter similarity (Traj), and then we propose a technique based on Fourier coefficients for symbolic-music analysis. We compare clusters built upon original values and Hertz, gaining hints on information found via proposed sonification. Possible future developments point to deeper connections between music theory, computer science, statistics, and precision medicine.