Recent advances in healthcare technologies have enabled the collection of large amounts of biological samples across various techniques and applications. In particular, over the past few years, Raman Spectroscopy (RS) has been successfully applied to the early-stage diagnosis of diseases. However, the inherent complexity and variability of spectra make manual analysis challenging, even for domain experts. For this reason, there has been exponential growth in the number of studies applying Machine Learning (ML) for automatic spectra classification. Despite their success, ML models are often considered black-box systems, meaning it is difficult to interpret the patterns they learn from the training data. To address this limitation, approaches in the field of eXplainable Artificial Intelligence (XAI) have been proposed. Among these, SHAP, a method based on cooperative game theory, is one of the most widely used. While SHAP provides strong theoretical guarantees and good empirical results, its computational complexity poses significant challenges for problems characterized by a large number of features. In this paper, we propose a novel XAI method inspired by SHAP and based on the concept of Quotient Game (QG) from cooperative game theory with coalition structure. Our approach is specifically designed to effectively handle spectral data characterized by high-dimensionality. We validate the effectiveness of our method using two real-world datasets of saliva Raman spectra: one for Covid-19 diagnosis and another for the identification of Parkinson’s and Alzheimer’s diseases. The explanations generated by our Quotient Game approach are compared with state-of-the-art methods through computational experiments.