Analysis of pattern electroretinogram signals of early primary open-angle glaucoma in discrete wavelet transform coefficients domain

Purpose: To evaluate discrete wavelet transform coefficients and identify descriptors of pattern electroretinogram (PERG) waveforms in order to determine PERG characteristics for optimizing the diagnosis of early primary open-angle glaucoma (POAG). Methods: Pattern electroretinogram was performed in 30 normal eyes and 30 eyes with primary open-angle glaucoma according to the ISCEV protocol. The check size was 0.8° and 16°, and the color was black/white in both groups. The results were analyzed in time domain (TD) and discrete wavelet transform (DWT) using the MATLAB software. The mean value, standard deviation, and relative energy of level 6 and 7 detail coefficients (d6, d7) and level 7 approximation coefficients (a7) of Daubechies 4 (db4), Daubechies 8 (db8), Symlet 5 (sym5), Symlet 7 (sym7), and Coiflet 5 (coif5) wavelets were calculated. In all the mentioned wavelets, DWT descriptors were extracted. Signals were reconstructed by inverse DWT. All data obtained by TD and DWT analyses were compared between the two groups. Results: In both check sizes, a significant attenuation of N95 amplitude was seen in the patient group. The relative energy of a7 of db8 increased significantly in the POAG group in the 0.8° check size. In larger check stimuli, the relative energy of d7 of coif5 decreased significantly and the standard deviation of d7 of sym7 increased markedly in glaucomatous patients (P < 0.05). In small stimuli, N95 descriptor (7N) of db8 had the highest value and showed a significant increase as compared to the POAG group. In the 16° check size, there was no significant difference. A strong correlation was seen between reconstructed signals and originals (r = 0.99). Conclusion: The DWT can quantify PERG responses more accurately. In agreement with TD and wavelet coefficients domain results, 7N of db8 decomposition can be used as a good indicator for early detection of POAG. © 2019, Springer Nature B.V