Abstract: A method for analyzing human brain physiology and pathology according to metabolic data obtained by non-invasive means. Metabolism is detected using in vivo proton nuclear magnetic resonance spectroscopy (.sup.1 H MRS) using pulse sequences that provide nuclear magnetic resonance signals from a living human brain. Factors which might decrease performance of artificial neuronal network analysis, e.g. the residual water signal, are eliminated from the nuclear magnetic resonance signal, e.g., by means of a singular value decomposition algorithm. Moreover, the artificial neuronal network analysis accepts as input all of the metabolite resonances, not just selected features, for analysis. The artificial neuronal network analysis yields metabolite concentrations and the nature of tissue type in focal lesions, such as tumors, stroke lesions, epileptic foci, traumatic scars, and also in more diffuse pathologies such as metabolic brain disorders as well as in post-therapeutic changes.