Abstract—This paper presents a framework to segment and extract key features automatically in Optical Coherence Tomography (OCT) scans. One of the main features to be detected is the Lamina Cribrosa (LC), which is an optic nerve head structure believed to play a crucial role in glaucomatous optic neuropathy. Detection of the LC aids in understanding pathogenesis and detection of glaucoma. Automatic segmentation allows a quick and objective way of identifying the LC. In previous work, LC segmentation has been manual; hence, the aim is to achieve automatic and accurate segmentation. Automatic detection is a novel approach, and very important as it provides an objective and fast way to identify the features. The method consists of three steps: automatic detection of the Bruch’s membrane opening, definition of LC Region of Interest (ROI), and feature detection in the ROI using local and inter-frame information. The best-fit curve representing the anterior LC was obtained by optimizing parameters to minimize the inter-frame gradient and local gradient change. The algorithm was applied to OCT images captured by Spectralis OCT machine (Heidelberg Engineering GmbH, Heidelberg, Germany). The results were compared and verified against manual segmentation of the key features, with a Root-Mean-Square (RMS) error of 9.89 and Dice coefficient of 0.74. The generally accurate results indicate that the approach is highly promising, and could potentially be expanded across detection in other image types.

Index Terms—automatic segmentation, Bruch’s membrane, feature detection, lamina cribrosa, optical coherence tomography

Cite: Mei Hui Tan, Sim Heng Ong, Sri Gowtham Thakku, Ching-Yu Cheng, Tin Aung, and Michael Girard, "Automatic Feature Extraction of Optical Coherence Tomography for Lamina Cribrosa Detection," Journal of Image and Graphics, Vol. 3, No. 2, pp. 102-106, December 2015. doi: 10.18178/joig.3.2.102-106