A Quantitative Analysis Of Fluid Related Abnormalities In Optical Coherence Tomography Images

Abstract

Blindness is one of the most common diseases amongst the newlinegreater population of India. In order to preserve proper vision, it has newlinebecome essential to have a regular check on the retinal conditions newlineperiodically, as age progresses. Optical Coherence Tomography is one newlinesuch imaging modality used in ophthalmology to diagnose the newlinedisorders in posterior region of the retina. With a notion to identify the newlinefluid based abnormalities in retina, namely, Cystoid Macular Edema, newlineChoroidal Neo Vascular Membrane and Macular Hole, the work newlineproposed aims at designing a computer based system that shall newlinecomment and interpret with reference to the input image and classify newlinethe same. Further to this, to provide a preliminary suggestion to the newlineend user, various general medical devices like ECG had imparted an newlineautomated optional interpretation system which shall comment on the newlineinput signal. Similarly, a scheme of interpretation has become more newlineconvincing for various imaging modalities also. Further to this, OCT newlinealso requires an automated expert system that could be integrated with newlinethe device itself, so as to interpret and leave some comments as and newlinewhen the images are captured. newlineThe proposed system takes the input image and performs newlinespeckle removal with the help of Homomorphic Wiener filters. Sobel newlineEdge Detection was identified to be an effective tool to isolate the newlinefluid filled boundaries in the preprocessed images. Zernike moments newlinewere found be wide varying feature for the application proposed. newlineThese features were extracted from the edge detected binary image newlineand were provided as input to classifiers. Taking three abnormalities in newlineviii newlineto consideration, K NN Classifiers and Random Forest Classifiers newlinehave equal metrics of performances. The Overall accuracy of the newlinesystem developed accounts to 91.65%, with Youden s Index of 0.83. newlineThe proposed system shall ease the technicians and also reduce the newlineworks of the ophthalmologists to a greater extent. This computerbased newlinesystem can also serve as an initial tool of analysis for various newlinedisorders.