Adjunct Methods for Alzheimer’s Disease Detection: A Review of Auditory Evoked Potentials

ASLR

ABR

ABR is a non-invasive technique that utilizes scalp electrodes for clinical evaluation and intraoperative detection of auditory function [70]. Unlike ECochG, ABR belongs to far-field ASLR and represents the summation of synchronous discharges from the auditory nerve to the brainstem in response to transient acoustic stimulation signals [54, 71]. Clinically, ABR plays a crucial role in detecting CANS abnormalities and is a time-locked neural response to sound recorded from the scalp using electroencephalogram (EEG) [72, 73]. ABR can help in evaluating the condition of patients with neurological diseases [71] and is suitable for studying precise time coding in the complete human auditory nerve system [70]. Ideally, ABR consists of seven waveforms denoted as Roman numerals I, II, III, IV, V, VI, and VII, which sequentially appear within 10 ms after stimulation, but only five waveforms or even fewer waveforms are commonly observed in human ABR (Fig. 1) [27, 70]. Based on the anatomical correlations with the main response components, wave I originates from the auditory nerve, wave II from the CN, wave III from the superior olive complex, wave IV from the LL, wave V from the IC, wave VI from the MGB, and wave VII from the AC [11, 27]. These waveform parameters, including threshold, amplitude, and latency, provide reliable estimates of auditory sensitivity, nerve reactivity and nerve conduction velocity in the CANS [5, 54]. Therefore, ABR is valuable in localizing and diagnosing lesions occurring in the CANS of individuals with AD [70, 74].

ABR is commonly elicited using either click stimulation or tone stimulation [75]. Click stimulation involves delivering rectangular electrical pulses with a duration of 50∼200μs to headphones or speakers, resulting in a broadband noise signal. The duration of click stimulation lasts for several milliseconds, and its frequency spectrum is wide, with the main energy concentration at 3000∼4000 Hz. Click stimulation is highly effective in inducing synchronized nerve impulses, leading to clear response waveforms, but it lacks frequency specificity [76]. On the other hand, tone stimulation provides instantaneous signals with specific frequencies, making it useful for assessing auditory sensitivity at different frequencies [77].

Animals are typically anesthetized through intraperitoneal injection before conducting ABR using an auditory electrophysiological processing system [78]. ABR amplitudes range from 1μV to 10μV and increase monotonically with sound pressure level [79]. The most commonly used animal model of AD is the amyloid precursor protein/presenilin 1 (APP/PS1) mice, which exhibit early-onset high-frequency hearing loss measured by ABR at 2 months of age compared to normal C57BL/6J mice [80]. In contrast, spatial learning defects in APP/PS1 mice emerge much later at 6 months of age [80]. At 3 months of age, APP/PS1 mice show a significant reduction or complete disappearance of ABR waves IV and V, prolonged latency of wave II, and prolonged interval between waves I and II, indicating the substantial impairment in the upper auditory centers of the IC, LL, and MGB in the brainstem [80]. Another AD animal model, 3xTg-AD mice, exhibits a notable loss of SGNs in the cochlea at 9 months of age, potentially attributed to the accumulation of phosphorylated tau protein and apoptosis-related protein in the cochlea with age [67]. Consequently, the ABR threshold significantly increases in 3xTg-AD mice at 9 months of age [67]. Similarly, 5xFAD mice, another AD animal model, experience a significantly elevated ABR threshold at 13∼14 months of age, accompanied by significant loss of cochlear hair cells at 15∼16 months of age [81]. BACE1^tm1Psa mice, a mouse model closely associated with AD, exhibit decreased amplitudes and increased latencies of waves I∼V, especially wave I, and disorganized and abnormally swollen SGNs [82]. ADNP-deficient mice, which share pathological features with AD individuals, display higher hearing thresholds and longer latencies in ABR click results, compared with normal mice of the same fetus, possibly linked to inner ear-auditory nerve-brain axis damage observed through immunohistochemistry and gene expression analysis [83]. A study on aging in gray mouse lemurs, proposing a model for early primate evolution and AD research, revealed a high variation in ABR thresholds among the old adult group, with one male even being deaf [79]. Early studies have demonstrated prolonged ABR latency in AD individuals despite having normal hearing thresholds [84, 85]. However, although ABR has demonstrated its advantages and usefulness in hearing assessment in AD animal models and a few clinical cases, its current implementation in clinics is limited. At present, ABR is primarily employed in clinical settings for infant hearing loss detection, and only a few clinical cases have utilized ABR as an adjunct method for AD detection [70]. Its clinical application still has a long way to go.

AMLR

Initially discovered as an AEP in 1958, AMLR has attracted considerable attention as an adjunct method to measure the neural integrity of the auditory thalamocortical pathway [75]. AMLR, when combined with ABR, can provide valuable information about components of the CANS, enabling differentiation between brainstem and thalamic cortex involvement [42, 55]. Similar to ABR, AMLR can be elicited using either click stimulation or tone stimulation, with the central zone electrode site typically yielding the most prominent response [75]. The absence of AMLR strongly indicates dysfunction in the CANS [75].

AMLR consists of four distinct waves commonly referred to as Na, Pa, Nb, and Pb (the same waveform as P50 in AERP) (Fig. 1) [49]. Most researchers agree that AMLR is generated from anatomical sites spanning from the IC to the AC [86]. Pa, believed to originate from Heschl’s gyrus in the primary AC [77, 87], reflects the auditory information processing time along the thalamo-cortical pathways and AC [25]. Pb, described as the most positive peak occurring between 40∼75 ms after the conditioning stimulus [49], is thought to be generated by pedunculopontine nucleus cells with cortical projections [85]. It serves as a measure of sensory gating, which is crucial for an individual’s ability to selectively attend to salient stimuli while ignoring redundant or trivial information, thereby protecting the brain from information overload [88]. Deficiencies in cognitive inhibition processes have been observed early in the clinical course of AD [89]. The inhibition of redundant information is a prerequisite for efficient cognitive processing and is believed to be modulated by prefrontal attentional networks [90]. Pb gating could potentially serve as a marker for inhibition deficits and, consequently, be valuable for prognosis estimation [90]. Previous studies have indicated that increased slow waves occurring as early as 50 ms may reflect neurophysiological consequences of neuropathology in mild cognitive impairment (MCI) [46], and Pb abnormalities suggest the dysfunction in the midbrain cholinergic cells in AD [12, 91], and using Pb amplitude as an adjunct method for the transformation of MCI into dementia has shown potential to enhance the sensitivity and specificity of prediction [92]. Brain structures associated with these waveforms include the thalamus, thalamic reticular nucleus, and temporal cortex [27], and their dysfunctions contribute to impaired sensory gating in AD individuals [93].

ALLR

N1, P2, N2

All auditory stimuli elicit an obligatory negative component known as N1, which typically manifests around 100 ms, followed by a subsequent positivity referred to as P2, occurring approximately 180 ms, and N2, occurring approximately 200 ms [12, 104]. N1 is believed to be produced by both the primary and secondary AC and reflects the selection and attention towards stimuli [106]. It represents the afferent nerve processing and functions related to perception, filtering, and short-term memory [107]. The amplitude of N1 is associated with the arousal level, active attention and short-term memory [104]. When the arousal and attention are high, leading to a greater number of activated neurons, the amplitude of N1 increases [104]. The latency of N1 is related to the initiation and speed of information processing [106]. In AD individuals, the number of activated neurons is often reduced, resulting in decreased responses, reduced alertness, impaired attention maintenance, and difficulties in short-term memory and memory retention [4, 108]. These difficulties in both AD individuals and animal models contribute to the reduced amplitude and prolonged latency of N1, which indicates impairment in the early processing of information [12, 109]. P2 can be a valuable tool for assessing of cognitive function [110–112]. Specifically, P2 can be elicited in response to rare tones using an active paradigm correlated with stress score [12], and can evaluate the ability to long-term cognitive deficits related to neurodegeneration [113]. Significantly smaller P2 amplitudes were obtained for participants with MCI than for cognitively normal older adults [111, 114]. N2 reflects the discrimination stage of stimulus analysis, and changes in N2 latency indicate the variations in the difficulty of discrimination [115]. The neural generators of N2 mainly include the frontal and superior temporal cortices, and N2 reflects the sensory cognition resulting from repetitive changes in auditory stimulation [116]. N2 latency may help identify individuals at risk of developing AD, while N2 amplitude could be valuable for monitoring longitudinal decline [116, 117]. Patients with cognitive impairment related to AD have shown delays in objective measurements utilizing N2 latency to distinguish perceive and classify auditory signals [118, 119]. Additionally, larger N2 amplitude and stronger activation of the left temporal lobe have been observed in the early stages of AD during information retrieval after long maintenance periods [112].

MMN

MMN can be elicited through auditory oddball detection paradigms and obtained by subtracting the AERP to the standard stimulus from that to the deviant stimulus, resulting in a frontal negativity at around 100∼200 ms [120–122]. Previous research has shown that the MMN initially originates from the primary AC in the temporal area and subsequently involves the secondary AC in the frontal area, with signals from the AC triggering frontal mechanisms associated with the attentional switching [122, 123]. Consequently, the impaired auditory sensory memory reflected by MMN in AD individuals may indicate reduced neural plasticity in the temporal lobe [120]. Furthermore, MMN subcomponents may also be generated in the thalamus and hippocampus [124]. MMN is closely linked to attentional processes, and its latency determines the response time to auditory changes [125]. Since it does not require the active attention or participation of the subjects, it can be helpful in functionally evaluating cognitive activities for AD individuals who may have difficulty cooperating during examinations [122].

Abnormalities in MMN amplitude and latency are closely associated with cognitive changes and declines observed in various neurological and neuropsychiatric conditions, as well as in normal aging [126]. MMN provides insights into the defects of sensory memory storage and automatic mismatch detection mechanisms, which may be used in the assessment of cholinergic dysfunction in AD [121]. Imaging studies in a patient with AD have revealed diffuse atrophy in the temporal, frontal and parietal lobes, and pathological examination have shown a decrease in nerve cells and neuronal vacuolation degeneration [122, 126]. These pathological and functional changes make it challenging for AD individuals to process new information. In such cases, standard stimuli fail to leave memory traces in the brain, and pathological neurons may not always maintain reactivity to deviation stimuli, which delays the encoding of deviation stimuli, prolongs the latency of MMN and reduces its amplitude [126].

P3

P3 has the longest history of clinical application among the AERP and has been the most widely used to evaluate the cognitive decline associated with various diseases affecting the CANS [101, 127–129]. It is a prominent positive AERP component, typically peaking around 300 ms after the onset of the stimulus [7, 130]. P3 reflects the attention and memory processes and is elicited when a patient is unexpectedly stimulated inconsistently, making it relevant to fundamental aspects of cognitive function [99, 131]. Moreover, P3 is independent of cultural and educational influences, making it a valuable, non-invasive tool for researching cognitive processes [102].

P3 accurately reflects the advanced integration function of the brain related to cognitive processes serving as an objective and sensitive electrophysiological index [132, 133]. It reflects cortical activity and likely represents simultaneous activity in multiple brain regions, including temporoparietal neocortical areas and higher limbic structures [94, 135]. Measurement of P3 latency may contribute to the objective detection of prodromal patients with AD and identify individuals with a high risk of AD several years before the clinical features of AD manifest [103, 136]. As AD progresses, a prolonged P3 latency has been assumed to be due to a subtle yet progressive cognitive decline [96, 138]. Numerous studies have reported prolonged P3 latency in AD individuals, indicating a decrease in the classification speed in functional attention-driven discrimination processing and task processing of AD individuals [92, 139]. P3 latency represents the process from stimulus reception to response, encompassing aspects such as the speed of stimulus identification and coding, comparison with the original information, correcting, and storage of stimuli, and the prolongation of P3 latency may be related to disordered information processing [140]. Additionally, the decrease in cholinergic activity in AD has an impact on P3 latency [102, 128]. P3 amplitude, considered as a measure of CANS, reflects the degree of effective resource mobilization during brain information processing, and is generally regarded as an indicator of resource allocation, particularly during working memory tests [141]. Typical P3 changes are shown in Fig. 2.

OPEN IN VIEWER

Fig. 2

P300 waveforms and parameters (mean and SD) from the three groups (Controls, MCI, and AD) recorded at Fz (anterior) and Pz (posterior). Lat stands for latency and Amp stands for amplitude. Fig. 2. is from [103].

P3 has two functionally different components: the earlier P3a, which is maximal over frontocentral regions, and the later P3b, which is maximal at posterior scalp locations [113, 142]. The auditory two-tone oddball task elicits either P3a or P3b, while the auditory three-tone oddball task simultaneously elicits both P3a and P3b waveforms [143]. Their peaks produced the most sensitive and reliable measures of the cognitive deficits associated with early-stage AD [97]. P3a is associated with attention mechanisms and the processing of novel stimuli [100], while P3b is more relevant to stimulus evaluation and decision-making processes. P3b is believed to be generated during memory storage operations involving the activation of temporoparietal areas, and serves as a mediator between perceptual analysis and response initiation, while also monitoring whether the decision to classify a stimulus is appropriately transformed into action [100, 139].

N4 and P6

N4 and P6 are the late components in AERP that play a role in semantic analysis. N4 is a negative wave observed at about 400 ms after stimulus onset and sensitive to semantic manipulations [144]. N4 may come from multiple generators in the parahippocampal gyrus, medial temporal structures near the hippocampus and amygdala, and lateral temporal region [145]. The amplitude of N4 tends to be larger when the semantic content are inconsistent [144]. P6 is a positive wave observed at about 600 ms after stimulus onset during both memory encoding and retrieval processes [44]. P6 can serve as an index for both memory encoding and retrieval [146]. The amplitude of P6 typically increases when previously presented words are successfully retrieved, while it decreases when words are repeated in a predictable, fixed context [44]. The inclusion of semantically congruous and incongruous word pairings within a repetition paradigm may potentially enhance the early detection or diagnosis of AD, enabling the assessment of both P6 and N4 [44]. Furthermore, in patients with MCI, reductions in either the P600 or N400 word repetition effect are associated with greater likelihood of subsequent transition to AD dementia [146].

Assessing ARHL could provide an avenue for early detection of AD

ARHL is closely associated with AD according to the recent hypotheses that the auditory afferent pathway can be regarded as a clinical biomarker and modifiable risk factor for AD [5, 148]. Auditory system dysfunction occurs in the precursor stage of AD, even before the appearance of clinical symptoms [24, 147]. In individuals with an increased risk of AD, abnormalities in neuronal processes detected by AEPs may precede the onset of clinical symptoms related to AD [114]. Specifically, dysfunction in CANS implicated in ARHL may occur prior to the manifestation of suspected clinical dementia associated with AD [5, 24].

The CANS plays a role in the pathology of AD, and histological changes within the CANS may contribute to cognitive function alterations associated with sensory decline [3]. While peripheral auditory function may not differ significantly from age-matched control group [65], dysfunction in the CANS becomes evident even in individuals with mild AD cases [27, 149]. It cannot be simply attributed to the confusion of speech-based cognitive tasks caused by hearing loss, as both AD and ARHL exhibit a highly specific and consistent pattern of degeneration [3, 148]. AD pathology involves the presence of amyloid-β plaques, tau protein aggregation and neurodegeneration, which are found not only in the AC of elderly AD individuals [150], but also in several key components of the CANS [18]. Structural abnormalities within the brainstem nerve bundle and nucleus of AD individuals may occur before observable neuron atrophy, indicating that the sensory dysfunction, particularly in hearing, may precede the onset of AD-related structural abnormalities [31, 151]. Previous studies have demonstrated the distribution of senile plaques and neurofibrillary tangles throughout the ventral nucleus of the MGB, the central nucleus of the IC and the AC in AD individuals [3, 8]. Even in the early stages of AD, significant neuronal loss and synaptic alterations occur in the MGB and IC [5, 152]. Additionally, auditory information processing disorders emerge early in AD individuals, as their information processing function in the AC are impaired, making it difficult for them to search for and encode the entered stimulation information. This degenerative sensory memory hinders the flow of information from preventing complete access to short-term memory and later higher-level processing, which can be sensitively reflected by AEPs [121]. Notably, the hippocampus is also implicated in the processing of sound, in which synaptic plasticity in AD is obviously impaired [6]. Eventually, CANS lesions in ARHL patients lead to changes in synaptic activity and the interruption of neural connectivity in the neurocognitive network, resulting in AD [14].

AEPs could be adjunct methods for early detection of AD

Given that AEP waveforms correspond to different aspects of the CANS, conducting hearing tests based on AEPs becomes essential for the early diagnosis of AD. AEPs have shown potential as a non-invasive, simple and repeatable technique to objectively examine assess the CANS, making them a potential screening tool for individuals at risk of AD [46, 153]. Unlike the visual evoked potentials, there is significant signal processing at each nucleus in the CANS [51], and pathological changes in any part of auditory nerve system, including both ascending and descending pathways, can result in abnormalities in AEPs [5]. For AD individuals, AEPs not only provide insight into CANS dysfunction but also higher-order auditory processing [66]. Moreover, the advantage of AEPs lies in their ability to evaluate patients who may not be able to cooperate with other AD detection methods. The AEP testing process requires minimal patient cooperation, as it does not necessitate specific responses or active participation from the patient. In contrast, traditional hearing detection methods, such as behavioral hearing threshold tests and speech perception tests, rely on specific responses like repeating test words, making them more demanding in terms of patient cooperation. AEPs can also serve as a means to monitor the progression of AD or evaluate the effectiveness of hearing treatments such as hearing aids and cochlear implants [22, 64]. By comparing the AEP results before and after treatment, doctors can accurately assess the effectiveness of treatment and make timely adjustments to the treatment plan.

More clinical data are crucial to further promote the utilization and advancement of AEPs

The existing data on AEPs are currently scattered and insufficient to establish a standardized experimental process, which limits the reliability of distinguishing healthy and cognitively impaired elderly people, especially in the early stage of AD [84]. Conducting a large number of clinical experiments is necessary to ensure the consistency, repeatability of experimental operations, and the quality and reliability of data. To deepen the analysis of the feasibility of ASLR, it is imperative to upgrade clinical hardware technology. Then more stable and accurate ECochG analysis will help to deepen the analysis of the AD influence on hearing from the perspective of peripheral hearing changes. Besides, clinical experiments with longer follow-up time are also needed to provide further support for the effectiveness of ABR, because some studies found no difference in ABR between the AD group and the healthy age-matched control group [46, 154], which may be attributed to the variations in disease severity and duration among AD individuals [9]. For both AMLR and ALLR, conducting more experiments can contribute to optimizing the selection and placement of electrodes and aid in the development of a better stimulation paradigm. Overall, addressing these limitations and gathering more clinical data will significantly contribute to the utilization and advancement of AEPs for AD detection and diagnosis.

Further optimization and hybrid application of AEPs

Further optimization of AEPs can be achieved through two main avenues: upgrading hardware and optimizing signal processing algorithms. Hardware optimization plays a crucial role in improving the signal-to-noise ratio and increasing the sampling rate. These enhancements result in easier-to-analyze signals, providing valuable insights into different neural activities. Optimizing signal processing algorithms is equally, as it improves the resolution of different frequency components or peaks and valleys observed in AEPs. This optimization can lead to a more accurate analysis of AEP characteristics. In particular, the development of artificial intelligence (AI) in the field of AEPs presents promising opportunities. Ingenious algorithm design utilizing AI has been shown to enhance the accuracy of AD diagnosis based on AEPs, even with the existing low-cost equipment [155, 156]. Additionally, AI can potentially reduce the equipment requirements and simplify operation of AEPs without compromising accuracy [156]. In the future, AI-based assessment of AEPs holds the potential to assist otolaryngologists or rehabilitation specialists in improving the differentiation between hearing and cognitive impairments and locating the lesion parts within CANS [157]. However, the availability of existing data sets in related studies is limited as mentioned above. A large-scale data set is needed to enhance the accuracy and stability of the algorithm to improve the sensitivity and specificity of AEPs in identifying ARHL patients at risk of AD [98, 136].

AEPs in predicting the future onset of AD is still in the early stages of clinical diagnosis. Currently, the main diagnostic criteria for AD detection still rely on cognitive and behavioral clinical evaluations, brain imaging examinations such as magnetic resonance imaging (MRI) or computed tomography (CT), and the detection of biomarkers [5, 147]. In addition to AEPs, there are several other sensory-based evoked potentials that exhibit potential as sensitive markers of early AD or other neurodegenerative diseases [112, 159]. But the hybrid application of AEPs with EEG, MRI, CT and other professional technologies may form the future development of a unique hybrid technology that can be further utilized to assess the risk of neurodegenerative diseases and enhance the accuracy of AD prediction [44, 160]. Integrating AEPs with psychological assessment methods such as neuropsychological tests based on auditory verbal learning can also aid in the clinical diagnosis of AD.