Sex Classification from Human Scent Using Image Interpretation of 2D Gas Chromatography-Mass Spectrometry Data

Two-dimensional gas chromatography coupled with time-of-flight mass spectrometry (GCxGC ToF-MS) is a powerful technique for analyzing complex chemical mixtures, capturing rich chemical information valuable for applications from environmental monitoring to medical diagnostics. One promising application is sex classification based on human scent, where subtle differences in chemical compounds can indicate biological sex. In this paper, we propose the first pattern recognition approach to sex classification that interprets raw GCxGC ToF-MS data as images, moving beyond traditional compound-based analysis. Furthermore, we introduce a new, publicly available dataset of GCxGC ToF-MS measurements specifically curated for this task. We evaluate our method on the proposed dataset, achieving state-of-the-art results of $\approx 95\%$ cross-validation accuracy with a dataset of $200$ identities, advancing the analysis of GCxGC ToF-MS data. The results show that applying computer vision techniques to chemical data analysis has significant potential for interdisciplinary research.