A novel method for automatic retinal detachment detection and estimation using ocular ultrasound image

This paper presents a novel method for automated detection of retinal detachment from ocular ultrasound image using digital image processing and computational techniques. Retinal detachment (RD) is an ocular emergency in which retina gets detached from the tissues lying underneath it and often requires immediate intervention to prevent rapid, irreversible vision loss. Direct fundoscopy and visual field testing are most common methods for the detection of RD. These methods are difficult to perform and they do not completely rule out retinal detachment. Generally, Ophthalmologists use ocular ultrasound to enhance their clinical acumen in detecting RD. Sometimes it is difficult to extract diagnostic features from ultrasound (USG) images due to its poor quality. Also, noise present in the image would cause misinterpretation during visual inspection;this demands development of intelligent and automated techniques for detection of retinal detachment. Further, the paper proposes a novel frame work for accurate and automatic retinal detachment using image processing techniques and mathematical analysis of detached area contour detected within the ocular globe. Furthermore, the estimation of diagnostic parameters, indicative of retinal detachment is also computed. Based on the mathematical analysis, three such parameters, percentage area of detached retina (PADR) compared to the ocular globe, angular width of detachment (α) and maximum radial distance of detachment to choroid layer beneath it (β), are calculated. These estimated parameters are very useful in determining the exact location and extent of retinal detachment. Results obtained through the proposed retinal detachment detection scheme are validated by the radiologist.