Autofluorescence spectroscopy for early diagnosis of cancer eye

Abstract

We report an in-vitro autofluorescence spectroscopic study of cow eye tissue to explore the applicability of the approach in discriminating early stage "cancer eye" from normal squamous eye tissues. Significant differences were observed in the autofluorescence signatures between the "cancer eye" and normal eye tissues. The spectral differences were quantified by employing a probability-based diagnostic algorithm developed based on recently formulated theory of Relevance Vector Machine (RVM), a Bayesian machine-learning framework of statistical pattern recognition. The algorithm provided sensitivity and specificity values of 97 ± 2% towards cancer for the training set data based on leave-one-out cross validation and a sensitivity of 97 ± 2% and a specificity of 99 ± 1% towards cancer for the independent validation set data. These results suggest that autofluorescence spectroscopy might prove to be a quantitative in-vivo diagnostic modality for early and accurate diagnosis of "cancer eye" in veterinary clinical setting, which would help improve ranch management from both economic and animal care standpoint.