Abstract
Abstract: :
Purpose: To demonstrate the effectiveness of a hyperspectral retinal imaging system for differentiating edematous macula from unaffected macula in a patient with clinically significant macular edema (CSME) by spectral signatures. Methods: A 33–year old male with a 15 year course of Type I diabetes mellitus and CSME confirmed by fluorescein angiography, and an age–matched control subject were dilated in one eye using 1% mydriacyl and 2.5% phenylephrine. Data were collected across the non–edematous regions, and if present, the edematous regions of the macula to produce 64 spectral bands simultaneously across the visible and near infrared region of the spectrum with a spectral resolution of 4nm. The same area was spatially imaged using a digital high resolution scene camera. Spectral signatures of clinically identifiable edema and unaffected macular tissue were analyzed using Principal Component Analysis (PCA) to identify features for classification. Results: Three different macula conditions or spectral phenotypes were identified from the data of the CSME subject. These were judged to represent the normal macular tissue, macula–CSME transitionzone, and definite CSME. PCA identified that only the first three principal components were required to categorize the tissue phenotypes. In addition to the identification of normal regions in CSME–associated retina, the normal fundus had a reduced variation in feature values suggesting an overlap between features. Conclusions: We have demonstrated the utility of a technique to characterize CSME in diabetic retinopathy based on hyperspectral data. Spectral feature variability may be related to reduced variability in the normal tissue composition compared to that of pathologic tissue. Further studies using this technology may demonstrate its utility to detect tissue changes pre–clinically. Disclaimer: This work was sponsored by the National Medical Technology Testbed, Inc. (NMTB) and the department of the Army, Cooperative Agreement Number (DAMD 17–97–2–7016). The content of the information in this work does not necessarily reflect the position or the policy of the government or NMTB. No official endorsement should be inferred.
Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • diabetic retinopathy