Altered retinal signaling in neurodevelopmental conditions such as autism spectrum disorder (ASD) (Ritvo et al., 1988; Constable et al., 2016, 2020b), schizophrenia and bipolar disorder (Hébert et al., 2010, 2015, 2020; Lavoie et al., 2014; Maziade et al., 2022) have been reported with differences between these groups and controls in time-domain parameters of the electroretinogram (ERG) waveform. Whilst this method is widely employed in cardiology (Addison, 2005; Meziani et al., 2013) and neurology (Bullmore et al., 2003; Faust et al., 2015), wavelet analysis has not been used previously to study time-frequency domain of the ERG in the neurodevelopmental disorders ASD and Attention Deficit Hyperactivity Disorder (ADHD). Discrete Wavelet Transform (DWT) analysis may provide biomarkers that facilitate stratification of these neurodevelopmental conditions (Molloy and Gallagher, 2021) which are known to share substantial genetic risk (Rommelse et al., 2010; Brainstorm Consortium et al., 2018, Cao et al., 2022).

Autism spectrum disorder and ADHD are the main neurodevelopmental disorders diagnosed in early childhood with a global prevalence estimated at approximately 1 and 3.4%, respectively (Polanczyk et al., 2015; Nevison et al., 2018). The co-occurrence of neurodevelopmental conditions is high with approximately 1:3 children with ASD also meeting diagnostic criteria for ADHD (Berenguer-Forner et al., 2015). ASD and ADHD often present as co-occurrence in children (Russell et al., 2014; Mansour et al., 2017) as well as other co-occurrences of conditions such as anxiety, epilepsy, and sleep disorders (Bougeard et al., 2021). Visual perception in ASD shows superiority in visual search (Constable et al., 2020a) and abnormal electrophysiological cortical differences in global motion perception (van der Hallen et al., 2019), motion onset (Constable et al., 2012), and coherence thresholds perception (Robertson et al., 2014). For reviews of sensory and visual perception in ASD see Dakin and Frith (2005) and Robertson and Baron-Cohen (2017). Individuals with ADHD also display differences in visual tasks and perception relating to visual attention (Zarka et al., 2021) and visual search (Mullane and Klein, 2008). Sensory processing problems are common in both ASD and ADHD children and a measure of sensory function using the ERG may help our understanding of the differences and similarities of these two groups (Dellapiazza et al., 2021).

The retina has three cell types connected in a vertical signaling pathway from the photoreceptors to the bipolar cells and then to the ganglion cells. Lateral neurons modify this path at two points: horizontal cells feedback to regulate the signal between photoreceptors and bipolar cells and amacrine cells link bipolar and ganglion cells (Masland, 2012). The retinal signal in response to brief flashes of light is captured as the ERG waveform. This is composed of an initial a-wave, a negative deflection originating from the hyperpolarization of the photoreceptors. The b-wave is a positive peak following the a-wave and is formed principally by the depolarization of bipolar cells. The cone and rod bipolar cells contribute to either the ON- or OFF- pathways within the retina (Kaneda, 2013). The ON-bipolar cells use slower metabotropic glutamate receptors and respond to an increase in retinal illumination, whilst- the OFF-bipolar cells utilize faster ionotropic glutamate receptors and respond when there is a decrease in retinal illumination (Severns and Johnson, 1993; Hanna and Calkins, 2006, 2007). The a-wave of the light-adapted ERG is shaped principally by the hyperpolarization of the cones but also has post receptoral contributions from bipolar cells (Bush and Sieving, 1994; Robson et al., 2003; Friedburg et al., 2004). Inhibitory pathways are formed by the horizontal and amacrine cells that utilize GABA and dopamine as the principal inhibitory neurotransmitters respectively (Diamond, 2017). The Oscillatory Potentials (OPs) are high frequency waves that appear on the b-wave and contribute to its amplitude and are initiated by the amacrine cells with some ganglion and bipolar cell modulation (Wachtmeister, 1980, 1981, 1998, 2001).

The DWT allows the energy to be determined within frequency bands at discrete time windows within the whole ERG waveform. The method was first applied to the ERG by Gauvin et al. (2014) who used a DWT derived from a continuous wavelet transform function to extract the energy associated within the frequency bands centered on: 20, 40, 80 and 160 Hz within the a- and b-wave time windows. The group then quantified the relative contributions of the DWT coefficients within each frequency band to specific sub-components of the ON- (20 Hz) and OFF- (40 Hz) pathway responses within the a- and b-waves (termed a20, a40, b20, b40) as well as characterizing the slow and fast OPs at 80 (op80) and 160 Hz (op160), respectively (Gauvin et al., 2014, 2015, 2016, 2017). This work demonstrated the fundamental strength of the DWT analysis in its power to quantify the discrete energies related to the ON- and OFF-pathways and the OPs. Our study aimed to explore the ability of the DWT analysis to identify features in the ERG waveform that could distinguish the neurodevelopmental conditions ASD and ADHD from a control group.

Ours is the first study to apply a DWT analysis of the ERG in individuals with a neurodevelopmental disorder. Our aim was to test the hypothesis that there would be underlying differences between ADHD and ASD in the light adapted ERG waveforms that we have previously reported as abnormal in ASD (Constable et al., 2016, 2020b). We did find a pattern of difference between the ASD and control groups OPs. Whilst some differences in the shape of the second light-adapted OP peak have been described previously in a small cohort of adults with ASD (Constable et al., 2016), we can now provide the first quantification of this component of the ERG in children with ASD. We propose that the reduced b-wave amplitude we found in association with ASD could be explained by reduced energy and contribution of the neural generators of the OPs. The OPs originate in the amacrine cells, which utilize dopamine as their main neurotransmitter (Wachtmeister, 1998, 2001). Genetic studies have suggested dopamine regulation plays a role in ASD (Liu et al., 2021; Pavăl and Micluţia, 2021). The characteristics of the coefficients op80, op160 and %OPs we observed, in relation to the b-wave, could be related to the amacrine cells and dopamine signaling, transport or storage (Wachtmeister and Dowling, 1978; Wachtmeister, 2001, 1981). Supportive evidence is provided by the observation that reduced retinal dopamine levels in Parkinson’s disease are associated with lower b-wave amplitudes (Jackson et al., 2012; Nowacka et al., 2015).

Our findings in respect of ADHD concerned the b20 and b40 coefficients of the ERG, which relate to the ON- and OFF-pathways. These use glutamate as the principal neurotransmitter. There appears to be greater energy in those coefficients than in controls. The a20 coefficient was increased in ADHD compared with controls at flash strengths from 0.40 to 1.20 log cd.s.m^–2 implying there is an ON-response difference between groups in this component. There is no direct involvement of the ON-pathway with the a-wave of the light-adapted ERG and the 20 Hz component is representative of the generalized on-response within the retina (Gauvin et al., 2015, 2017). In contrast, the a40 or OFF-pathway component to the a-wave did not differentiate these two groups at flash strengths greater than 0.60 log photopic cd.s.m^–2 suggesting there is a relatively normal OFF-pathway response within the time window of the a-wave in children with ADHD. The post-receptor OFF-pathway normally contributes the amplitude of light-adapted a-wave (Bush and Sieving, 1994; Robson et al., 2003; Friedburg et al., 2004). The elevated energy in DWT coefficients that we observed may reflect alterations in bipolar cell functions that contribute to the b-wave. These implicate glutamate signaling and or transport/storage as the underlying cause, and glutamic gene function is associated with hyperactivity and impulsivity in ADHD (Naaijen et al., 2017). A polymorphism in metabotropic glutamate receptors predisposes to ADHD in the Chinese Han population (Zhang et al., 2021).

The %OPs were not significantly different between ADHD and control for flash strengths lower than 1.20 log photopic cd.s.m^–2 (p > 0.11) with the main coefficients contributing to the elevated b-wave in ADHD being b20 and b40. In contrast, in keeping with the reduced op80 and op160 coefficients in ASD the %OPs were significantly lower for the ASD group compared to the control group for all flash strengths greater than 0.4 log photopic cd.s.m^–2 (p ≥ 0.005) supporting the reduced contribution of the OPs to the overall b-wave amplitude in ASD.

Evidence from mouse models also supports the clinical findings here. A recent study using the BTBR inbred mouse strain as an ASD model showed reductions in the b-wave response under dark and light adapted conditions (Cheng et al., 2020) as well as. Studies of an ADHD mouse model in which novel DA transporter had been knocked out have indicated ERG b-wave amplitudes are increased under light-adapted conditions, which is consistent with our findings (Dai et al., 2017). Offspring of mice exposed to valproic acid as an ASD model also show reduced dark-adapted ERG responses (Guimarães-Souza et al., 2019) in support of previous findings of adults and children (Ritvo et al., 1988; Constable et al., 2016) and a deficit in ON-pathway retinal function. Currently no dark-adapted ERG studies have been performed in ADHD and this would be of interest to determine if the elevated ON-pathway response was also present under a different state of retinal adaptation.

The pathophysiology of ADHD and ASD remains a conundrum. There are no biomarkers for either condition, and we know multiple genetic and environmental factors contribute to the phenotypes (Geurts et al., 2013; Cooper et al., 2014; Kohls et al., 2014; Ronald et al., 2014; van Steijn et al., 2014; Naaijen et al., 2017). We provide some tentative evidence for neurophysiological changes that not only differentiate both conditions from typically developing children, but also evidence that they can be distinguished from each other based on ERG characteristics. Our findings suggest an elevation in the overall energy in the ERG within the b-wave and OPs in ADHD, which is consistent with reports of greater background retinal noise in this group (Bubl et al., 2015; Lee et al., 2022). In contrast, we report a reduction in the OPs contribution to the b-wave in ASD. We suggest this could account for the reduction in the b-wave amplitude previously reported in those with ASD, under dark and light adapted conditions (Ritvo et al., 1988; Constable et al., 2016, 2020b,2021; Lee et al., 2022).

Discrete wavelet transform analysis has been applied to other clinical conditions. For instance, the multifocal ERG and the pattern ERG reveal a positive association with primary open angle glaucoma (Brandao et al., 2017; Hassankarimi et al., 2019). In congenital stationary night blindness, DWT analysis of the ERG and multifocal ERG highlighted the known absence of the ON-pathway involvement in the waveforms (Dorfman et al., 2020). A DWT analysis of low contrast pattern reversal visual evoked potentials at different spatial frequencies found the DWT coefficients showed greater significant differences than time domain or fast Fourier transform parameters for describing the evoked potentials (Hassankarimi et al., 2020). In cardiac disease, wavelet analysis has been used extensively for analysis of the ECG and cardiac arrhythmia (Rahul and Sharma, 2021; Li et al., 2022). It has also been used to analyze EEG signals in epilepsy (Zarei and Asl, 2021) and Parkinson’s disease (Liu et al., 2017). The application of retinal signal analysis may provide a new marker for the classification of neurodevelopmental and neurodegenerative disorders given the broad heterogeneity and the lack of sufficient objective clinical and neuroscientific evidence within current classifications (Molloy and Gallagher, 2021). There is a need for a different approach that can support more reliable diagnoses, prognoses, and better targeted treatments. The DWT methodology has the potential to provide a “biotype” based on retinal signal analysis (Clementz et al., 2016; Cuthbert, 2020) and contribute to this field. Signal analysis of the ERG waveform may also provide more subtle functional insights into the structural changes within the retina which have been documented in longitudinal studies of Alzheimer’s or Parkinson’s disease using retinal imaging techniques (Kashani et al., 2021). It may also provide a sensitive method to monitor the effects of clinical trials of pharmacological and gene therapy to manage retinal and ophthalmic diseases (Yu-Wai-Man et al., 2020; Maguire et al., 2021; Gajendran et al., 2022).

Analysis of the ERG waveform using wavelet functions [as well as other signal analysis techniques such as Variable Frequency Complex Demodulation (VFCDM), (Chon et al., 2009), or Functional Data Analysis (Calle-Saldarriaga et al., 2021)] could provide better characterization of the heterogeneity inherent within the neurodevelopmental disorders that share common clinical traits (Thapar et al., 2017; Molloy and Gallagher, 2021). Of these, VFCDM can provide one of the highest time-frequency resolutions (Chon et al., 2009) and has been used to analyze other physiological signals (Wang et al., 2006; Siu et al., 2009, Posada-Quintero et al., 2016; Hossain et al., 2021). VFCDM uses a bank of low-pass filters to decompose the signal into a suite of band-limited signals. These can estimate instantaneous amplitude, frequency, and phase within each frequency band with higher resolution than DWT. VFCDM may offer an additional process to unlock the hidden signals within the ERG of neurological and retinal conditions.

This study is the first to apply a wavelet transform approach to analyzing clinical waveforms in neurodevelopmental conditions using the ERG. The method allows an objective description of the ON- and OFF- pathways as well as the contribution of the OPs to the overall ERG signal. A combination of time domain (b-wave amplitude) and time-frequency domain parameters (b20, b40, op80, op160) may assist with developing classification models of ADHD and ASD in the future, as well as providing a simple method to extract information on the underlying contributions of the ON- and OFF- pathways without the need for modeling luminance-response functions across several flash strengths (Hamilton et al., 2007; Hébert et al., 2017; Constable et al., 2020b). We speculate that the wavelet approach may in future aid in the study of retinal neurophysiology in schizophrenia and bipolar disorder (Hébert et al., 2015, 2020) as well as neurodegenerative conditions such as Parkinson’s disease (Garcia-Martin et al., 2014; Nowacka et al., 2015). Further work will be required to establish abnormalities in these retinal signals are specific to neurodevelopmental disorders and what effects may be observed in the case of co-occurrence of other conditions (Silverstein and Thompson, 2020).

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://doi.org/10.25451/flinders.17712347.v3.

This study was reviewed and approved by the Flinders University Human Research Ethics Committee and the South East Scotland Research Ethics Committee in the United Kingdom. Written informed consent to participate in this study for those under 16 years of age was provided by the participants’ legal guardian/next of kin.

PC wrote the first draft and collected ERG recordings with IL and DT. FM-R conducted the statistical analysis. DS contributed to the main discussion and clinical assessments. PC and MG performed the DWT analysis. All authors contributed equally to the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.