Horizontal visibility graph analysis of melodic skylines in musical stimuli used in epilepsy research

Music has been explored as a non-invasive adjunctive intervention in drug-resistant epilepsy, yet the structural properties that differentiate musical pieces used in this context remain poorly quantified. Mozart's Sonata for Two Pianos in D major (K. 448) has repeatedly been associated with reductions in interictal epileptiform discharges, but the specific musical organization underlying these reports is still unclear. Here, we provide a descriptive, corpus-limited structural characterization of six compositions previously used as stimuli and/or comparators in epilepsy research, treating them as case studies with heterogeneous reported outcomes rather than as a clinical efficacy trial. Starting from score-derived symbolic data in the Musical Instrument Digital Interface format, we apply an explicit symbolic projection by extracting melodic skylines, defined as the leading contour formed by the highest active pitches. We then map skylines into Horizontal Visibility Graphs and compute graph-theoretical and information-theoretic descriptors on fixed-length, non-overlapping windows (W=100 skyline events), reporting within-piece variability as mean±standard deviation. A reduced-feature principal component analysis reveals a dominant structural axis contrasting global network integration (higher cohesion and shorter characteristic distances) with structural fragmentation (longer path lengths and larger diameters). Within this feature space, K. 448 occupies the integrated extreme, whereas Wagner's Lohengrin Prelude, used as an experimental comparator in the epilepsy literature, lies at the fragmented extreme. These results provide a reproducible framework for mapping symbolic musical organization onto complex-network signatures and generate quantitative hypotheses for future mechanistic work combining score-based analysis with electrophysiological recordings.