Purchase this article with an account.
D. Tudor, V. Kajic, B. Povazay, S. Rey, B. Hofer, A. Unterhuber, B. Hermann, W. Drexler, J. E. Morgan; Identification of the Optical Signature of Neuronal Programmed Cell Death by Ultra High Resolution Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2010;51(13):4804.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Neurons undergo prolonged periods of compromise and structural change prior to the activation of cell death pathways, identification of this period of compromise, and thus a therapeutic window, could allow timely interventions leading to a possible reversal in neuronal damage. The development of ultrahigh resolution optical coherence tomography (UHR-OCT) allows tissue to be imaged in vitro with an image resolution better than 2 µm. In this project we aim for the first time to identify the apoptotic signature of cells using UHR-OCT.
RGC-5 cells were cultured in DMEM with 10% fetal calf serum (FCS), 4 mM Glutamine, 100 U/ml penicillin and 100 mg/ml streptomycin in a humidified atmosphere with 5% CO2 at 37°C. Cells were seeded onto glass coverslips and maintained for 48-72hrs, then cultured in serum free DMEM growth medium, containing 1µm Staurosporine for 24hrs to induce apoptosis. Live cell images from healthy and apoptotic cells were obtained at a sampling 1024x466x1024 voxel at 20 Mvx/s, at 800nm central wavelength and a bandwidth of 230nm. Images were analyzed using ImageJ to generate 3D representations of cells within the region of interest, a 30x400 µm segment.Datasets were first preprocessed to remove artifacts along with histogram equalization. Texture analysis was then performed using 64 dimensional feature space where each dataset was projected as one point. A Gaussian mixture model was then used to form two clusters, one for healthy and one for apoptotic cells. To classify a new dataset the Mahalanobis distance was used to determine the distance to each of the clusters and the minimum distance determines what class the new dataset belongs to.
To evaluate the model, a leave-one-out test was performed on each 80 datasets, i.e. one dataset was left out and the clusters were then formed on the rest of the data, the left out dataset was then classified. Using this method of analysis, we were able to classify healthy and pathologic RGC-5 populations with 100% classification accuracy.
UHR-OCT coupled with texture analysis is a promising technique of the detection of neuronal damage prior to cell death. This technique may have potential for the in vivo detection of neuronal damage, avoiding the use of ligands or fluorophores.
This PDF is available to Subscribers Only