Background. There is very little knowledge in scientific literature about methods to classify intervertebral disc tissues in real time. This work is dedicated towards this purpose.
Materials and methods. During intervertebral disc hernia operation in Republican Vilnius University Hospital pieces of intervertebral disc tissue were removed as part of conventional procedure. Optical properties of these samples were investigated using fluorescence spectroscopy method using no artificial or additional fluorophores. 1085 fluorescence spectra were registered from specimens of intervertebral disc hernia tissues. These specimens were classified to 3 medical groups according to localization of tissue (D – specimen from inner part of a disc, S – free sequester, P – herniation from subligamentous space). The fluorescence was excited using 355nm radiation of a microlaser STA-01-TH (Standa, ltd.).
Analysis. Registered spectra were normalized to maxim peak value. Noise of fluorescence signal was reduced using the smoothing procedure. Next, spectra analysis followed. There was made an assumption, that the whole fluorescence spectrum of a specimen is a sum of fluorescence spectra of several fluorescent materials (fluorophores) that the sample is consisted of. The real spectra of these materials are unknown. Thus the registered spectra were decomposed to smaller spectra, called the components, using multi curve resolution (MCR) alternating least square (ALS) algorithm [1],[2]. Each spectrum was described by the intensities of all extracted components. These intensities were used to train an artificial network (ANN) to identify spectra of medical (D, S, P) groups.
Results. During this analysis more than 75% of spectra were identified correctly as belonging to appropriate medical group.
Conclusion. Combination of spectra decomposition to components using MCR-ALS algorithm and implementation of an ANN seems a promising direction for analysis of intervertebral disc hernia fluorescence spectra.