A new technique to improve glaucoma discrimination was more effective than conventional data analysis at differentiating between normal eyes and those with suspected glaucoma. It also performed at least as well as the conventional method at distinguishing between normal and glaucomatous eyes, according to Dr Juan Xu, PhD, who described the technique in a presentation at the annual meeting of the Association for Research in Vision and Ophthalmology.

The new technique, developed by the Glaucoma Imaging Group at the University of Pittsburgh Medical Center (UPMC) Eye Center under the direction of Dr Joel Schuman, chairman, is based on superpixel machineclassifier analysis of three-dimensional (3D) spectral-domain optical coherence tomography (SDOCT). While the faster scanning speed of SDOCT allows 3D volume scanning of the retinal layers and enhances early detection of glaucoma and monitoring of disease, these structural measurements do not take full advantage of the 3D dataset, Dr Xu noted.

Only 512 sampling points out of 40000 are used in the conventional retinal nerve fiber layer (RNFL) analysis, and signs of pathologic damage may be missed.

"Quantitatively summarizing the full 3D dataset is still a fundamental challenge in structural analysis and progression monitoring," said Dr Xu, research associate, UPMC Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, University of Pittsburgh School of Medicine. "Therefore, the purpose of this study was to develop a 3D SDOCT data analysis technique to analyse the full dataset [quantitatively] using [the] super-pixel machine-classifier [method]."