Algorithm for LIF spectra evaluation in cervical lesions.

A8

Two steps algorithm for CIN discrimination by LIF point monitoring was created. Light induced fluorescence point monitoring fitted to live cervical tissue diagnostics in 42 patients. Ultraviolet (337 nm) laser excited fluorescence spectra in the live cervical tissue analyzed. The recorded 598 spectra in these samples were put in matrix and cross correlation with the histology data and clinical features was counted. The correlation coefficient with its level of significance (alpha=0,050) was established. Building a model for predicting medical features from fluorescence spectra principal component regression PCr) analysis performed Posterior probability of being inflamed correct classification renders, sensitivity 74, specificity 69, 3, area under curve AUC=72, 4%. PCr method best discriminated pathology group "CIN I and inflammation" ( AUC=75%) related to fluorescence emission in short wave region, 5 PC were included in analysis. Next step of algorithm: group "CIN I and inflammation" exclusion. 10-12 PC inclusion in next step evaluation gives better classification result for CIN II discrimination (>70AUC). Immunity related changes in live tissue influence Laser induced fluorescence method discrimination value in cervical precancer lesion diagnostics.