The analysis of high-resolution, proton, Nuclear Magnetic Resonance (NMR) spectrometry can be obscured by the adoption of black-box algorithms.
We seek an intermediate representation able to furnish a more communicative interface between human expert and machine.
The representation is based on spectral peaks.
These are coded and used to compile a dictionary for all spectral traces which is then used to further transform the trace data into a format, know as the Bag of Peaks, useful for classification.
Our pilot study, of Type I diabetes among Sardinian children, demonstrates the efficacy of Bag of Peaks descriptors over those of a standard PCA.
The talk begins with an overview of NMR spectrometry and the algorithmic techniques used to analyze the traces it produces.
We then motivate our particular approach before illustrating the obtained results.
