INTEGRATED SYSTEM AND METHOD OF ACOUSTIC SPECTROSCOPY FOR THE ANALYSIS, CHARACTERIZATION AND CLASSIFICATION OF MATTER ASSISTED BY ARTIFICIAL INTELLIGENCE

Abstract

The present invention describes an integrated acoustic spectroscopy system and method for the analysis, characterization, and classification of organic, inorganic, and biological matter, assisted by supervised artificial intelligence, through the analysis of samples obtained from specific entities. The system integrates a computing unit (2) with a graphical interface (1), a function generator (3), a signal processing device (4), and an acoustic coupling clamp (5) with coaxial transducers. As an application example, glass samples (from inorganic entities), culture media (from organic entities), and cell culture lines (from biological entities) were successfully characterized and classified.The method involves capturing a digitized acoustic signature from a sample, generating a multidimensional data hierarchy that ranges from time-domain signals to high-density spectrograms obtained by Continuous Wavelet Transform (CWT) and post-processing vectors for classification. Using a 2% rescaling, standardization, and dimensionality reduction (UMAP) chain, an acoustic signature is extracted and fed into a Support Vector Machine (SVM) model optimized by Bayesian inference. The invention is notable for a management and synchronization module that allows for incremental retraining of the model through manual labeling in the dynamic repository. The system enables the differentiation of healthy and pathological cell phenotypes with a Matthews Correlation Coefficient (MCC) greater than 0.94 in less than a minute, optimizing the analysis, characterization and classification of matter without dependence on reagents or complex infrastructures, with outstanding application in cancer screening.