Improving the reliability of the pattern electroretinogram with adaptive filtering.

Purpose: Electrophysiological methods enable exact layer by layer diagnosis of the retina and help to answer frequent diagnostic ambiguities between macular and optic nerve diseases. Methods: New multifocal techniques enable topographic representation of the retinal function and can provide accurate diagnosis of macular diseases in early stages. Current trends with emphasis on non‐invasive recording of EOG, ERG, pattern ERG, VEP and multifocal testing, with regards to conventional perimetry, microperimetry and scanning laser ophthalmoscope imaging including autofluorescence of the RPE will be reviewed in representative clinical cases. The importance of standardisation on the performing of all tests according to the recommendations of the International Society for Clinical Electrophysiology of Vision (ISCEV) is emphasised. Results: Global retinal function is well assessed by flash evoked ERGs, but these may be normal in macular disease. PERG and multifocal ERG in combination with autofluorescence imaging may delineate macular diseases in very early stages. In such cases, visual loss may, due to the resulting delay in VEP, erroneously be attributed to optic nerve disease. On the other hand, optic nerve diseases usually do not affect multifocal ERG but can affect PERG, especially its N95 component. Combination of reduced N95 component and delay in VEP is strongly suggestive for optic nerve or ganglion cell disease in which autofluorescence imaging would usually be normal. Conclusions: By judging the cause of visual loss, combination of morphological features by RPE autofluorescence with electrophysiological and psychophysical methods usually leads to correct diagnosis.

The photopic negative response (PhNR) may be useful as a tool to monitor longitudinal change in retinal ganglion cell (RGC) function. The goal was to assess PhNR test-retest reliability, and to estimate the amount of change between tests that is likely to be statistically significant for an individual test subject. Photopic electroretinograms (ERGs) were recorded from 49 visually normal subjects (mean age, 38.9 years; range, 21-72 years). Signals were acquired using Dawson-Trick-Litzkow (DTL) electrodes in response to red stimulus at four flash energies (0.5, 1, 2.25, 3 cd·s/m2) on a blue background (10 cd/m2). The PhNR amplitude was recorded from prestimulus baseline to trough (BT), prestimulus baseline to fixed time point (BF), and b-wave peak to trough (PT). The ratio of baseline PhNR to b-wave amplitude (BT/b-wave) was calculated. Reliability was assessed using the intraclass correlation coefficient (ICC2,1) and coefficient of repeatability (CoR). Flash energy of 1.00 cd·s/m2 produced reliable, well-defined traces. At this stimulus, the a- and b-wave amplitudes were reproduced with moderate reliability (ICC, 0.62; CoR%, 90.0%; and ICC, 0.74; CoR%, 54.3%; respectively). For PhNR, the order from most to least reliable measurement was: PT (ICC, 0.64; CoR%, 59.1%), BT (ICC, 0.40; CoR%, 148.3%), and BF (ICC, 0.22; CoR%, 166.1%). The BT/b-wave did not improve reliability (ICC, 0.37; CoR%, 181.5). The b-wave peak-to-PhNR trough amplitude produced the most reliable measurement. A relatively large magnitude of change in PhNR amplitude is required to make clinical inferences about changes in RGC function. Refinement to the technique of acquisition and/or processing of the PhNR is recommended to improve reliability.

We recorded pattern electroretinograms, scotopic threshold responses, oscillatory potentials and ganzfeld flash electroretinograms in patients with glaucoma or other optic nerve diseases and in cats with inner retinal damage caused by intravitreal injections of kainic acid. In both studies, the scotopic b-wave and the scotopic threshold responses were normal but the oscillatory potentials and pattern electroretinograms were not. The photopic b-wave was also often reduced in patients with scotopic oscillatory potential reduction, and the reduction was proportionate to the oscillatory potential change. Oscillatory potentials were as frequently reduced as pattern electroretinograms in both patient groups, and in the few cases where only one response was reduced, there was no bias toward either measure. In cats, the effects of intravitreal injection of various doses of kainic acid on the retina were evaluated electrophysiologically, and structural damage was assessed histologically. After 25 nmol of kainic acid, the pattern electroretinograms and oscillatory potentials were reduced but neither the b-waves nor the scotopic threshold responses, were affected. Histologic studies of retinas after this dose showed swollen dendrites that were restricted to the outer part (off-sublamina) of the inner plexiform layer. Serial semithin sections indicated that most, if not all, of the swelling was confined to dendrites of large ganglion cells. Our results indicate that the size and sensitivity of the oscillatory potential response may have a role in the diagnosis and management of early glaucoma and optic nerve disease, and that the photopic electroretinogram may give similar information.

Dynamic mechanical allodynia is traditionally induced by manual brushing of the skin. Brushing force and speed have been shown to influence the intensity of brush-evoked pain. There are still limited data available with respect to the optimal stroke number, length, force, angle and speed. Therefore, an automated brushing device (ABD) was developed, for which brushing angle and speed could be controlled to enable quantitative assessment of dynamic mechanical allodynia. To compare the ABD with manual brushing using capsaicin-induced allodynia, and to investigate the role of stroke angle and speed on pain intensity. Experimental dynamic mechanical allodynia was induced by an intradermal injection of capsaicin (100 µg) into the volar forearm of 12 healthy, male volunteers. Dynamic mechanical allodynia was rated on a 10 cm visual analogue scale (VAS) after each set of strokes at angles of 30°, 60° and 90° with speeds of 17 mm⁄s, 21 mm⁄s and 25 mm⁄s for each angle. A two-way ANOVA with repeated measures was performed to assess the influence of brushing parameters. To evaluate test-retest reliability, Bland-Altman 95% limits of agreement, including a coefficient of repeatability and an intraclass correlation coefficient (ICC), were determined. The angle and speed exhibited a significant impact on pain intensity (P<0.001 and P<0.015, respectively). Post hoc analysis showed that the highest pain intensity was recorded with an angle of 30° regardless of brushing speed. The ABD demonstrated superior test-retest reliability (coefficient of repeatability = 1.9 VAS; ICC=0.91) compared with manual brushing (coefficient of repeatability = 2.8 VAS; ICC=0.80; P<0.05). The most reliable combination of parameters (coefficient of repeatability = 1.3 VAS; ICC=0.97) was an angle of 60° and a speed of 21 mm⁄s. A controlled, automatic brushing method can be used for quantitative investigations of allodynic reactions, and is more reliable for quantitative assessment of dynamic mechanical allodynia compared with traditional manual brushing.

Background Ultra-long scan depth OCT can achieve imaging of full range of human ocular anterior segment.However, the measurement of the dimension of anterior segment from the OCT image with high speed and precision is a challenge at present.The software of automatic data processing is still lack in analyzing spectral domain OCT. Objective This study was to perform the automatic biometry and data processing of human ocular anterior segment OCT image by using self-developed automatic detection software and evaluate the accuracy and repeatability of this method. Methods Twenty eyes of 10 normal subjects were included in Eye Hospital of Wenzhou Medical University from June to July 2013.The OCT image of anterior eye segments were obtained with custom-made ultra-long scan depth OCT under the informed consent.An automatic software algorithm was developed for the biometric measurement on these OCT images, including boundary segmentation, image registration and optical correction of OCT images.The boundary segmentation algorithm utilized the axial gradient information of OCT images and the shortest path search principal based on the dynamic programming to optimize edge finding.Central corneal thickness (CCT), anterior chamber depth (ACD), pupil diameter (PD), lens thickness (LT), radius of lens anterior curvatures (LAC) and radius of lens of posterior curvatures (LPC) were automatically and manually measured, and the validity of automatic detection algorithm was assessed by calculating the intraclass correlation coefficient (ICC) between the automatic and manual measurements, and the repeatability was validated by calculating the coefficient of repeatability (COR) between repeated measurement.This study was approved by the Ethic Committee of Wenzhou Medical University and informed consent was obtained from all subjects. Results There were no significant differences in the results of CCT, ACD, PD, LT, LAC and LPC between the automatic and manual measurements (P=0.205, 0.167, 0.285, 0.127, 0.102, 0.074). The results were consistent between automatic and manual measurements (all at ICC>0.75). The repeated measurement values were consistent in CCT, ACD and LT in both automatic and manual modes (all at ICC>0.75). The reproducibilities of automatic biometry in PD and LAC (ICC=0.793, 0.872; COR=2.90, 5.79) were better than those of manual mode (ICC=0.631, 0.579; COR=5.62, 10.46); while the reproducibility of automatic biometry in LPC (ICC=0.663; COR=6.17) was lower than that of manual mode (ICC=0.794, COR=4.79). Conclusions Self-developed automatic detection software appears to be accurate and repeatable in measuring dimension of spectral domain OCT images.This automatic software algorithm can be used for the biometry and monitor of human ocular anterior segment. Key words: Biometry/methods; Tomography, optical coherence/methods; Automatic data processing; Anterior eye segment/anatomyh Reproducibility

Transient pattern electroretinograms were recorded in 12 normal young volunteers with the use of 12' and 30' checkerboard stimuli and 1 c/deg and 2.5 c/deg vertical gratings. The intrasession repeatability of pattern electroretinograms was determined by employing alternative statistics to the frequently used correlation coefficients. Comparisons were made with the repeatability of cone electroretinograms elicited by ganzfeld stimulation. The coefficient of repeatability was substantially lower (better) for pattern electroretinogram b-wave implicit times than for a-wave implicit times and b-wave amplitudes, the latter having high coefficients of repeatability. Cone electroretinograms had better repeatability with respect to all three parameters. Digital filtering using a bandpass of 1-40 Hz was applied to the original pattern electroretinograms recorded with a bandpass of 1-250 Hz. Although peak definition was generally enhanced, repeatability did not show marked improvement. Differences between broadband analogue-filtered pattern electroretinograms and the same responses with high-frequency noise removed by digital filtering need to be taken into account. The coefficient of repeatability statistic appears to facilitate the assessment of pattern electroretinograms and permits the comparison of the repeatability of both implicit time and amplitude parameters irrespective of absolute values. The implications of the relatively poor repeatability of pattern electroretinogram b-wave amplitudes for clinical applications need to be considered.

Previous studies have shown that the onset of high-contrast, fast reversing patterned stimuli induces rapid blood flow increase in retinal vessels in association with slow changes of the steady-state pattern electroretinogram (PERG) signal. We tested the hypothesis that adaptive PERG changes of normal controls differed from those of glaucoma suspects and patients with early manifest glaucoma. Subjects were 42 glaucoma suspects (Standard Automated Perimetry-MD -0.89±1.8 dB), 22 early manifest glaucoma (MD -2.12±2.4 dB) with visual acuity of ≥20/20 and 16 age-matched normal controls from a previous study. The PERG signal was sampled every ~15 seconds over 4 minutes in response to gratings (1.6 cyc/degree, 100% contrast) reversing 16.28 times/s. Amplitude/phase values of successive PERG samples were fitted with a nonparametric locally weighted polynomial regression smoothing function to retrieve the initial and final values and calculate their difference (δ) and the residual SD around the fitted function. The magnitude of PERG adaptive change compared to random variability was calculated as log10 of percentage coefficient of variation (CoV)=100×residual SDr÷δ. Grand-average PERGs were also obtained by averaging all samples of the same series. The grand-average PERG amplitude [analysis of variance (ANOVA), P=0.02], but not phase (ANOVA, P=0.63), decreased with increasing severity of disease. Adaptive changes [log10 (CoV)] of PERG amplitude were not significantly associated with disease severity (ANOVA, P=0.27) but adaptive changes [log10 (CoV)] of PERG phase were (ANOVA, P=0.037; linear trend, P=0.011). The steady-state PERG signal displayed slow adaptive changes over time that could be isolated from random variability. PERG adaptive changes differed from those of grand-average PERGs (corresponding the standard steady-state PERG), thus representing a new source of biological information about retinal ganglion cell function that may have potential in the study of glaucoma and optic nerve diseases.

In this study we show how the Pattern ERG (PERG) can be used to distinguish between optic nerve and retinal disease. Records from eyes with RBN and delayed visual evoked responses are compared with those recorded from the normal fellow eyes. In optic nerve disease there is a selective reduction of the later negative component of the PERG. PERGs were also recorded from patients with mild diabetic retinopathy. These were divided into three groups on the basis of the delay in their Visual Evoked Potential (VEP). The amplitude of both positive and negative components of the PERG fall with increasing severity of the disease thus showing that the abnormally long delay in VEP found in some cases is due to retinal disease rather than optic nerve disease.

The objective of these studies was to assess the reproducibility of (i) short-term food intake (FI) and subjective appetite following a glucose preload, (ii) ventilation threshold (VT) and subjective appetite after short-duration exercise, and (iii) body composition assessed by bioelectrical impedance (BIA). On two separate weekend mornings, boys (n = 11; aged 9-14 years) received drinks containing 50 g glucose made up to 250 mL with water 2 h after a standardized breakfast. FI from a pizza meal was measured 30 min later. Subjective appetite was measured before and after the glucose loads and the VT measures. VTs were measured on 2 weekday evenings, 1 week apart. BIA was measured during the FI assessment sessions. Short-term FI after the glucose preload was highly reproducible. Mean energy intake was 925 +/- 139 kcal on the first day and 988 +/- 147 kcal on the second day (coefficient of repeatability (CR) = 259 kcal; intra-class correlation coefficient (ICC) = 0.96). Moderate reproducibility of the average appetite score was found at 30 min (CR = 24 mm; ICC = 0.82). Subjective appetite was increased similarly by short-duration exercise on both days (CR = 19 mm). Absolute VT was more highly reproducible (CR = 359 mL O2.min(-1), ICC = 0.85) than VT expressed on the basis of body weight (CR = 8.0 mL O2.kg(-1).min(-1), ICC = 0.59). Fat mass (FM) estimated from BIA was highly reproducible (CR = 2.7 kg, ICC = 0.95), but underestimated FM compared with skinfolds. In conclusion, FI and subjective appetite scores in response to glucose preloads, VT, subjective appetite after short-duration exercise, and estimates of FM from BIA are reproducible in boys.

Adaptive filters play an important role in the fields related to digital signal processing and communication, such as system identification, noise cancellation, channel equalization, and beamforming. In practical applications, the computational complexity of an adaptive filter is an important consideration. The Least Mean Square (LMS) algorithm is widely used because of its low computational complexity ($O(N)$) and simplicity in implementation. The least squares algorithms, such as Recursive Least Squares (RLS), Conjugate Gradient (CG), and Euclidean Direction Search (EDS), can converge faster and have lower steady-state mean square error (MSE) than LMS. However, their high computational complexity ($O(N^2)$) makes them unsuitable for many real-time applications. A well-known approach to controlling computational complexity is applying partial update (PU) method to adaptive filters. A partial update method can reduce the adaptive algorithm complexity by updating part of the weight vector instead of the entire vector or by updating part of the time. In the literature, there are only a few analyses of these partial update adaptive filter algorithms. Most analyses are based on partial update LMS and its variants. Only a few papers have addressed partial update RLS and Affine Projection (AP). Therefore, analyses for PU least-squares adaptive filter algorithms are necessary and meaningful. This monograph mostly focuses on the analyses of the partial update least-squares adaptive filter algorithms. Basic partial update methods are applied to adaptive filter algorithms including Least Squares CMA (LSCMA), EDS, and CG. The PU methods are also applied to CMA1-2 and NCMA to compare with the performance of the LSCMA. Mathematical derivation and performance analysis are provided including convergence condition, steady-state mean and mean-square performance for a time-invariant system. The steady-state mean and mean-square performance are also presented for a time-varying system. Computational complexity is calculated for each adaptive filter algorithm. Numerical examples are shown to compare the computational complexity of the PU adaptive filters with the full-update filters. Computer simulation examples, including system identification and channel equalization, are used to demonstrate the mathematical analysis and show the performance of PU adaptive filter algorithms. They also show the convergence performance of PU adaptive filters. The performance is compared between the original adaptive filter algorithms and different partial-update methods. The performance is also compared among similar PU least-squares adaptive filter algorithms, such as PU RLS, PU CG, and PU EDS. In addition to the generic applications of system identification and channel equalization, two special applications of using partial update adaptive filters are also presented. One application uses PU adaptive filters to detect Global System for Mobile Communication (GSM) signals in a local GSM system using the Open Base Transceiver Station (OpenBTS) and Asterisk Private Branch Exchange (PBX). The other application uses PU adaptive filters to do image compression in a system combining hyperspectral image compression and classification.

Objective: Non-invasive navigation techniques have recently been developed to determine mechanical femorotibial alignment (MFTA) in extension. The primary aim of this study was to evaluate the precision and accuracy of an image-free navigation system with new software designed to provide multiple kinematic measurements of the knee. The secondary aim was to test two types of strap material used to attach optical trackers to the lower limb.Methods: Seventy-two registrations were carried out on 6 intact embalmed cadaveric specimens (mean age: 77.8 ± 12 years). A validated fabric strap, bone screws and novel rubber strap were used to secure the passive tracker baseplate for four full experiments with each knee. The MFTA angle was measured under the conditions of no applied stress, valgus stress, and varus stress. These measurements were carried out at full extension and at 30°, 40°, 50° and 60° of flexion. Intraclass correlation coefficients, repeatability coefficients, and limits of agreement (LOA) were used to convey precision and agreement in measuring MFTA with respect to each of the independent variables, i.e., degree of flexion, applied coronal stress, and method of tracker fixation. Based on the current literature, a repeatability coefficient and LOA of ≤3° were deemed acceptable.Results: The mean fixed flexion for the 6 specimens was 12.8° (range: 6–20°). The mean repeatability coefficient measuring MFTA in extension with screws or fabric strapping of the baseplate was ≤2°, compared to 2.3° using rubber strapping. When flexing the knee, MFTA measurements taken using screws or fabric straps remained precise (repeatability coefficient ≤3°) throughout the tested range of flexion (12.8–60°); however, using rubber straps, the repeatability coefficient was >3° beyond 50° flexion. In general, applying a varus/valgus stress while measuring MFTA decreased precision beyond 40° flexion. Using fabric strapping, excellent repeatability (coefficient ≤2°) was observed until 40° flexion; however, beyond 50° flexion, the repeatability coefficient was >3°. As was the case with precision, agreement between the invasive and non-invasive systems was satisfactory in extension and worsened with flexion. Mean limits of agreement between the invasive and non-invasive system using fabric strapping to assess MFTA were 3° (range: 2.3–3.8°) with no stress applied and 3.9° (range: 2.8–5.2°) with varus and valgus stress. Using rubber strapping, the corresponding values were 4.4° (range: 2.8–8.5°) with no stress applied, 5.5° (range: 3.3–9.0°) with varus stress, and 5.6° (range: 3.3–11.9°) with valgus stress.Discussion: Acceptable precision and accuracy may be possible when measuring knee kinematics in early flexion using a non-invasive system; however, we do not believe passive trackers should be mounted with rubber strapping such as was used in this study. Flexing the knee appears to decrease the precision and accuracy of the system. The functions of this new software using image-free navigation technology have many potential clinical applications, including assessment of bony and soft tissue deformity, pre-operative planning, and post-operative evaluation, as well as in further pure research comparing kinematics of the normal and pathological knee.

To assess the repeatability of Goldmann-correlated intraocular pressure (IOPG), corneal-compensated IOP (IOPCC), corneal hysteresis (CH), and the corneal resistance factor (CRF) obtained with the Ocular Response Analyzer (ORA) in normal eyes and to determine whether any differences in corneal biomechanical parameters or their repeatability exist between the sexes. A prospective observational study assessing 100 normal adults (50 men and 50 women; median age, 54.5 years). Comparison of ORA parameters measured in both eyes in three sets of four consecutive readings by one examiner within a 30-minute period. The mean values of the ORA parameters assessed, the intraclass correlation coefficient (ICC) and the coefficient of repeatability (CR), were as follows for the right eye (n = 100): IOPG, 16.2 ± 3.3 mm Hg (ICC, 0.96; CR, 6.37); IOPCC, 16.2 ± 3.2 (ICC, 0.94; CR, 6.29); CRF, 10.9 ± 1.9 (ICC, 0.91; CR, 3.62); CH, 10.6 ± 1.7 (ICC, 0.94; CR, 3.37). The ICC between the readings was excellent (>0.9) for all the ORA parameters in both sexes, with the exception of one group (female CRF ICC, 0.86). Coefficient of repeatability of the instrument satisfies the British Standards Institution criteria for repeatability. There were no significant differences between corneal biomechanical factors and IOP results for men and women. Corneal biomechanical parameters measured by the ORA provide repeatable results in normal eyes, with no significant difference between the sexes.

Classification of macular and optic nerve disease by principal component analysis

Reproducibility of Pattern Electroretinogram in Glaucoma Patients with a Range of Severity of Disease with the New Glaucoma Paradigm

Adaptive filters are frequently employed in many applications in which the statistics of the underlying signals are either unknown a priori or slowly time-varying. Adaptive filtering algorithms are usually expected to have fast convergence speed, low computational complexity and high robustness to numerical problems and outlier interference. Many researchers have invested enormous efforts in deriving new algorithms with the above properties and analyzing their convergence behaviors. The latter is even more complicated due to the mathematical manipulations involved. Following the same guideline, in this dissertation we study a set of efficient adaptive transversal filtering algorithms and their convergence performance analysis. The development of the new algorithms and the establishment of the effective analytical framework are based on three important modeling approaches. (1) Noise modeling approach. By modeling the outliner impulsive noise as contaminated Gaussian distributed, we study the normalized least mean M-estimate (NLMM), transform domain NLMM (TD-NLMM) and partial update NLMM (PU-NLMM) algorithms which are more robust to impulsive noise than their conventional normalized least mean square (NLMS)-based counterparts. Complete convergence analyses of these algorithms are provided to interpret the underlying principles behind their performances. (2) Input modeling approach. By modeling the input signal as a low-order autoregressive process, the fast LMS/Newton algorithm can reduce the computational complexity of the traditional Newton-type algorithm while retaining its improved convergence speed. We propose two improved fast LMS/Newton algorithms. One is the block exact fast LMS/Newton algorithm which is mathematically equivalent to the original algorithm but has a significantly reduced complexity. The other is the robust fast LMM/Newton algorithm which is derived through the noise modeling approach used in (1). Moreover, we also develop a Newton-type algorithm with a uniform structure. It can realize flexible performance-complexity tradeoff and has the potential to be incorporated with the certain input modeling approach to achieve fast convergence performance with low complexity. (3) Channel modeling approach. By exploiting the sparse feature of the system channel encountered in vast applications, the generalized proportionate NLMS (GP-NLMS) algorithm possesses a faster initial convergence and tracking speed. Our proposed generalized proportionate stepsize (GPS)-fast LMS/Newton algorithm combines the advantages of the GP-NLMS and the fast LMS/Newton algorithms and exhibits a superior overall convergence and tracking performance. In addition, based on the GP-NLMS algorithm, another variable forgetting factor QR decomposition-based recursive least M-estimate (RLM) (VFF QR-RLM) algorithm is proposed. It has both an improved numerical stability and faster overall convergence and tracking speed than the conventional RLM algorithm using constant forgetting factor. All the proposed algorithms and the corresponding convergence analysis were tested through extensive computer simulations and good agreements between the theoretical predictions and the experimental results were observed. In general, the algorithms proposed in this dissertation have addressed some of the key problems in adaptive filtering algorithm design with the aid of three modeling approaches. They should form an algorithm set whose constituting components are intrinsically connected together and can potentially be utilized for various applications either individually or in a combination.