This study is basically related to this aspect of the stone structure characterization, since a new most robust methodology for the characterisation and classification of kidney stones has been developed trying to avoid expert dependence. More than 200 samples of renal calculi, including all main types of calculi, were analysed by means of stereoscopic microscopy and SEM, as shown in the literature3,4,so as to get a conventional classification of the stones, which served as reference classification. The use of Near Infrared Spectroscopy coupled to the hyperspectral imaging technique proved to be able to show the spectral characteristics for every compound. Moreover, the chemometric analysis of the data allowed the selection of some regions in the spectra which have stronger classification capabilities. The data was used for creating a model, using some considered pure samples, by means of Artificial Neural Networks. This computational model showed a really good performance for the classification of the main groups of kidney stones, including those which are actually mixtures of some compounds. Furthermore, the main advantage this model exhibits is the faster and more confident classification of the stones, compared to the conventional methodologies.
NIR-Hyperspectral Imaging and Artificial Neural Networks for the Characterisation of Renal Calculi / Francisco, Blanco; Montserrat López, Mesas; Serranti, Silvia; Bonifazi, Giuseppe; Josef, Havel; Manuel, Valiente. - ELETTRONICO. - 13:(2011), pp. 1-1. (Intervento presentato al convegno European Geosciences Union General Assembly 2011 tenutosi a Vienna, Austria nel 03-08 April 2011).
NIR-Hyperspectral Imaging and Artificial Neural Networks for the Characterisation of Renal Calculi
SERRANTI, Silvia;BONIFAZI, Giuseppe;
2011
Abstract
This study is basically related to this aspect of the stone structure characterization, since a new most robust methodology for the characterisation and classification of kidney stones has been developed trying to avoid expert dependence. More than 200 samples of renal calculi, including all main types of calculi, were analysed by means of stereoscopic microscopy and SEM, as shown in the literature3,4,so as to get a conventional classification of the stones, which served as reference classification. The use of Near Infrared Spectroscopy coupled to the hyperspectral imaging technique proved to be able to show the spectral characteristics for every compound. Moreover, the chemometric analysis of the data allowed the selection of some regions in the spectra which have stronger classification capabilities. The data was used for creating a model, using some considered pure samples, by means of Artificial Neural Networks. This computational model showed a really good performance for the classification of the main groups of kidney stones, including those which are actually mixtures of some compounds. Furthermore, the main advantage this model exhibits is the faster and more confident classification of the stones, compared to the conventional methodologies.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.