Graefe's Arch Clin Exp Ophthalmol (1989)227:21 25 Graefe's Archive for Clinical and Experimental Ophthalmology © Springer-Verlag 1989 In vivo thickness of the human detached retina by ultrasonic signal processing* Gloria Wu 1, Ronald H. Silverman 1, D. Jackson Coleman 1, and Frederic L. Lizzi 2 1 Department of Ophthalmology, Cornell University Medical College, and 2 Riverside Research Institute, New York, NY 10036, USA Abstract. Fourier operations on digitized radio-frequency (rf) data can provide the most sensitive measure of the thickness of thin, deterministic structures such as mem- branes. After deconvolution against the system transfer function, the cepstrum and the analytic signal magnitude can provide measures of membrane thickness on the order of a half-wavelength. A total of 19 patients with rhegmato- genous retinal detachments were scanned using a diagnostic ultrasound system adapted for analog to digital conversion of rf data. Data were analyzed using the signal processing techniques described above to measure retinal thickness. Recent retinal detachments tended to be significantly thicker than normal attached retinas, and thickness de- creased with the age of detachment. A statistically signifi- cant correlation was found to exist between retinal thickness and the duration of detachment. Introduction The axial resolution of a diagnostic ultrasound system is determined by a number of factors, the most important of which is transducer characteristics. Wavelength is the most obvious factor influencing resolving power, but band- width and focal characteristics are involved as well. Band- width defines the duration of the wave packet emitted when the transducer is pulsed and hence the axial length of the wave packet, R. Focal characteristics determine the lateral extent of the pulse, W. Together these factors define a vol- ume, proportional to RW 2, within which interference phe- nomena may occur. Ambiguity therefore exists as to the precise point of origin of echoes within such a volume. Based on these considerations, desirable transducer charac- terics should include high center frequency, broad band- width, and weak focusing, since this combination minimizes RW 2. Conventional A-scan display devices are designed for biometric measurements on scales of tens or hundreds of wavelengths. At such scales, biometric ambiguities on the scales discussed above may be tolerable. Some applications, however, may require biometric accuracy on the order of * This work was supported in part by NIH grants EY-03183 and EY-01212, the Dyson Foundation, and Research to Prevent Blind- ness Offprint requests to: D. Jackson Coleman, 1300 York Avenue, New York, NY 10021, USA a wavelength. In such cases, considerations of transducer characteristics and signal processing assume great impor- tance. In deconvolution, spectra are divided by the calibration signal (normalization) to expand the bandwidth (as speci- fied, e.g., at its half-power points); this process improves the inherent resolution of the echo signals. Two techniques can then be used to process further the deconvolved spectra. Cepstrum analysis [10] measures the thickness of determin- istic structures based on an analysis of the constructive and destructive interference pattern seen in normalized power spectra. Analytic signal magnitude [6] is computed to realize optimal envelope detection; i.e., one which does not signifi- cantly reduce the inherent resolution of the deconvolved signal. To compute the analytic signal, negative frequency components of the normalized Fast Fourier Transform (FFT) of the rf signal are set to zero. When returned to the spatial domain by an inverse FFT, the analytic signal magnitude constitutes a high-resolution A-scan whose peaks locate the margins of the measured structure. In this paper, we consider the case of in vivo ultrasonic measurement of the thickness of human detached retina and the relationship between retinal thickness and the age of detachment. The histologic characteristics of human retina in retinal detachment have been incompletely catalogued because of the paucity of histologic specimens. Our understanding of the retina in the pathologic process of retinal detachment has therefore been derived from the study of experimental retinal detachment in animals. Experimental evidence indicates that retinal detachment results in decreased thickness of the retina. In the owl mon- key, Machemer [9] described the diffuse loosening of the inner layers of the detached retina and cystoid space forma- tion in the first 3 days. By the 7th day of detachment the inner nuclear layer had significantly degenerated; the outer nuclear layer was reduced to one or two cells. Subsequently, there was significant atrophy in the number of photorecep- tors. The inner segments became short and thick, the outer segments almost disappeared. In experimental retinal detachment in the cat [3], retinal necrosis increased as a function of the time following de- tachment. After detachment, there was a significant loss of cells in the outer nuclear layer and migration of photore- ceptor cell bodies into the subretinal space. There was also degeneration of synaptic terminals in the outer plexiform layer by 2 weeks after detachment.