Cancer detection based on Raman spectra super-paramagnetic clustering

• The clustering of Raman spectra of serum sample from cancer patients is analyzed.
• The super-paramagnetic clustering (SPC) method to classify Raman spectra is proposed.
• By identifying the most natural clusters, SPC method allows discriminating the control and cancer patients.
• In leukemia case, SPC detects a hierarchical structure allowing the identification of the patient leukemia type.
• The goal of this study is to apply a model of statistical physics to classify spectra allowing the design of a cancer detection method.

The clustering of Raman spectra of serum sample is analyzed using the super-paramagnetic clustering technique based in the Potts spin model. We investigated the clustering of biochemical networks by using Raman data that define edge lengths in the network, and where the interactions are functions of the Raman spectra’s individual band intensities. For this study, we used two groups of 58 and 102 control Raman spectra and the intensities of 160, 150 and 42 Raman spectra of serum samples from breast and cervical cancer and leukemia patients, respectively. The spectra were collected from patients from different hospitals from Mexico. By using super-paramagnetic clustering technique, we identified the most natural and compact clusters allowing us to discriminate the control and cancer patients. A special interest was the leukemia case where its nearly hierarchical observed structure allowed the identification of the patients’s leukemia type. The goal of this study is to apply a model of statistical physics, as the super-paramagnetic, to find these natural clusters that allow us to design a cancer detection method. To the best of our knowledge, this is the first report of preliminary results evaluating the usefulness of super-paramagnetic clustering in the discipline of spectroscopy where it is used for classification of spectra.