Differential geometric approach for automated eardrum thickness measurement in OCT image processing

Abstract

We propose a method for the automated in vivo thickness measurement of the human eardrum in OCT imaging. The method is based on extraction curvilinear structures using eigenvectors and eigenvalues of smoothing image Hessian (Steger method). This allows for extracting the central line of the eardrum and calculating the normal to each point. The latter circumstance makes it possible to find the proper thickness of an arbitrarily oriented eardrum having a complex shape. The algorithm works correctly despite the presence of other image elements complicating the central line and borders extraction process (presence of an effusion or structures with bright borders-bones). The algorithm does not require large computing power and can be applied to real time image processing. In addition, a comparison of the effectiveness of the proposed method with a Canny edge detector is given.