Hypothesis Paper · Unlisted · Version 1.0 · June 2026
Subclinical Multisensory Integration Decline as an Early Biomarker in Alzheimer's Disease
The Sensory Integration Reserve Hypothesis and Its Implications for Digital Assessment.
Part of the NeuroFlex Research Series. See also Paper I — The Dynamic Cognitive Reserve Hypothesis. Status: pre-publication draft. Domain: Sensory Neuroscience · Alzheimer's Disease · Digital Biomarkers.
Alzheimer's disease (AD) is increasingly recognised as a multisystem neurological disorder whose clinical expression extends well beyond episodic memory loss to encompass changes across sensory processing, motor function, and higher-order perceptual integration. While individual sensory deficits — particularly olfactory loss and central auditory processing disruption — have been studied as prodromal AD markers, the domain of multisensory integration (MSI) remains substantially underexplored as a source of early biomarkers.
We propose the Sensory Integration Reserve Hypothesis (SIRH), which holds that the brain's capacity to efficiently combine signals across sensory modalities constitutes a distinct reserve function — Sensory Integration Reserve (SIR) — that may be compromised by AD neuropathology before conventional cognitive deficits become detectable. The SIRH predicts that subclinical degradation of cross-modal binding, temporal alignment of multisensory signals, and audiovisual speech integration will precede measurable cognitive decline and will track with underlying neuropathological load.
We further propose that modern smartphones provide a sufficient and practically accessible platform for the passive and semi-active assessment of MSI function, and that longitudinal digital measurement of MSI — combining visual, auditory, and haptic stimuli in reaction-time and adaptation paradigms — could constitute a novel class of non-invasive digital biomarkers for early AD detection and monitoring. We outline the theoretical grounding, falsifiable predictions, and measurement architecture required to evaluate these claims.
- Introduction
- Known Sensory Changes in Alzheimer's Disease
- Beyond Single Senses: The Multisensory Integration Framework
- The Sensory Integration Reserve Hypothesis
- Falsifiable Predictions
- Digital Measurement Architecture
- NeuroFlex as a Sensory Integration Monitoring Platform
- NeuroFlex Research Programme: A Coherent Line
- Discussion
- Limitations
- Conclusion
- References
Introduction
For much of its clinical history, Alzheimer's disease has been framed primarily as a disorder of memory. This framing, while not wrong, is incomplete. AD is a neurodegenerative process affecting broad cortical and subcortical networks, and its functional consequences extend to sensory processing, perceptual organisation, spatial reasoning, and the orchestration of information from multiple sensory channels simultaneously [1, 2]. The singular focus on episodic memory as the cardinal clinical feature has arguably delayed recognition of a wider set of early functional changes that may manifest substantially earlier in the disease course.
Among the most promising and underutilised early markers are those related not to the performance of a single sensory system, but to the brain's capacity to integrate information from multiple sensory systems in a coordinated and temporally precise manner. This capacity — multisensory integration (MSI) — relies on distributed parietal, temporal, and frontal networks whose connectivity and function are known to be disrupted by AD neuropathology in its earliest stages [3].
The present paper proposes the Sensory Integration Reserve Hypothesis (SIRH), a theoretical framework that positions MSI as a measurable reserve function — analogous in structure to the cognitive reserve framework, but operating at the level of cross-modal perceptual binding rather than higher-order cognition. We argue that subclinical MSI decline may constitute an early, practical, and digitally measurable biomarker of AD onset, and that its systematic assessment has been neglected largely due to the methodological demands of laboratory MSI paradigms — demands that modern smartphone hardware is increasingly capable of meeting.
This paper is the second in a planned series advancing NeuroFlex's theoretical research programme. The first paper established the Dynamic Cognitive Reserve Hypothesis [4], arguing that cognitive resilience in AD depends on ongoing behavioural engagement rather than solely on lifetime accumulation. The present paper extends this logic into the sensory domain, proposing that a parallel form of reserve — one concerned with integrative perceptual processing — operates by analogous principles and is similarly susceptible to early, subclinical disruption.
Known Sensory Changes in Alzheimer's Disease
Before introducing the novel claims of the SIRH, it is important to establish what is already known about sensory system involvement in AD. Sensory deficits in this condition are neither peripheral nor coincidental; many arise from central neural changes related to the primary disease process and are detectable, in several modalities, substantially before the onset of clinical dementia.
Most research on sensory deficits in AD examines changes within a single modality. The SIRH is concerned with a different and arguably more sensitive phenomenon: changes in how the brain combines information from multiple modalities simultaneously. This integrative capacity draws on distributed associative networks that are among the earliest and most severely affected by AD neuropathology [3].
Beyond Single Senses: The Multisensory Integration Framework
In everyday experience, the brain does not process sensory inputs in isolation. When a person watches someone speak, the brain combines visual information about lip movements with the acoustic signal of the voice to produce a unified percept — a process so robust that discrepancies between the two modalities generate striking perceptual illusions (the McGurk effect [13]). When we catch a ball, the brain simultaneously integrates visual, proprioceptive, and vestibular information to compute the ball's trajectory and prepare a motor response. This cross-modal binding is not a passive summation of unimodal signals; it is an active, computationally demanding process mediated by specific cortical regions — the superior temporal sulcus, the posterior parietal cortex, and prefrontal areas that regulate attentional modulation of sensory processing [3, 14].
MSI has two properties particularly relevant to its use as an early AD biomarker. First, it is highly sensitive to the integrity of the temporoparietal junction and superior temporal sulcus — regions that are among the earliest to show cortical thinning and metabolic decline in the preclinical and prodromal phases of AD [15]. Second, it is measurable with high precision using simple reaction-time paradigms: the benefits of multisensory redundancy (faster and more accurate responses to bimodal than unimodal stimuli) and cross-modal conflicts (slower and less accurate responses when signals are asynchronous or discrepant) provide quantitative indices of integration efficiency that are sensitive to subtle neural changes [16].
Subclinical AD disruption would appear as slower binding, reduced cross-modal benefit, increased sensitivity to temporal asynchrony, and impaired adaptation — measurable in isolation, before unimodal performance shows detectable change.
The temporal binding window (TBW) — the range of stimulus onset asynchronies within which the brain accepts signals from different modalities as belonging to a single event — has been proposed as a particularly sensitive measure of MSI efficiency [17]. A widened TBW, requiring a longer alignment between visual and auditory signals before the brain perceives them as synchronous, has been documented in normal ageing [18]; its behaviour in preclinical and prodromal AD is an open empirical question that the SIRH directly addresses.
The Sensory Integration Reserve Hypothesis
4.1 Core Proposition
Subclinical degradation of multisensory integration — the brain's capacity to efficiently combine concurrent signals from visual, auditory, and haptic channels into a unified, temporally coherent percept — constitutes a measurable and early functional consequence of Alzheimer's disease neuropathology that precedes the onset of conventionally detectable cognitive impairment.
This degradation reflects compromise of a distinct reserve capacity — Sensory Integration Reserve (SIR) — that buffers against neuropathological disruption to the associative and heteromodal cortical networks mediating cross-modal binding. Individuals with higher SIR will show preserved multisensory function at greater neuropathological burden; its loss will permit earlier clinical expression of MSI dysfunction, providing a biomarker signal that is both earlier and mechanistically distinct from those obtainable through unimodal sensory testing or standard neuropsychological assessment.
4.2 Sensory Integration Reserve (SIR)
By analogy with the cognitive reserve framework [19, 20], we propose that individuals differ systematically in their capacity to maintain MSI function against neuropathological disruption. This capacity — SIR — may be shaped by lifetime experience with multisensory environments (musical training, multilingualism, visuomotor skill learning, and socially rich communication settings are all candidate SIR-building experiences [21, 22]) and by ongoing engagement with tasks that demand cross-modal coordination.
The parallel with the Dynamic Cognitive Reserve Hypothesis is deliberate. Just as we argue in Paper I that cognitive reserve has both a structural component (accumulated over a lifetime) and a functional component (maintained by current engagement), we propose here that SIR similarly has a structural component — the organised connectivity of heteromodal associative cortex — and a functional component: the ongoing exercise of cross-modal binding through active engagement with multisensory environments. Under this view, individuals who remain in communication-rich, musically active, or visuomotor-demanding environments may maintain higher SIR expression, independently of their structural reserve.
4.3 Candidate Neural Mechanisms
Efficient MSI depends on a network of heteromodal association areas — the superior temporal sulcus (STS), the temporoparietal junction (TPJ), and posterior parietal cortex (PPC) — as well as on long-range synchronisation between early sensory cortices that enables temporal binding [3, 14]. Structural and functional MRI studies in early AD and MCI have documented:
- Early grey matter loss in the superior temporal sulcus and TPJ [15]
- Reduced functional connectivity between visual and auditory association cortices [23]
- Disrupted default mode network — posterior hubs of which overlap substantially with heteromodal association cortex involved in sensory integration [24]
- Elevated amyloid burden in posterior association cortices in the preclinical phase [25]
Predictive coding theory provides an additional mechanistic angle [26]: MSI can be understood as the brain generating predictions about expected cross-modal contingencies and adjusting them on the basis of incoming evidence. Disruption to the precision-weighting of sensory prediction errors — a known consequence of cholinergic deficits in AD — would directly impair temporal binding and cross-modal conflict resolution, producing measurable changes in MSI performance before overt cognitive deficits become evident.
Falsifiable Predictions
The SIRH generates the following testable predictions, each distinguishing it from the null hypothesis that MSI changes in AD are secondary to unimodal sensory or cognitive deficits:
| # | Prediction | How to Test |
|---|---|---|
| P1 | MSI efficiency measures (cross-modal reaction time benefit, temporal binding window width, audiovisual congruency effects) will be reduced in preclinical and MCI populations relative to age-matched controls, independently of peripheral sensory thresholds and standard cognitive test scores. | Cohort study comparing MSI performance between biomarker-confirmed preclinical AD, MCI, and healthy control groups, with co-measurement of unimodal sensory thresholds and standard neuropsychological battery. |
| P2 | Longitudinal MSI decline will predict conversion from MCI to AD dementia at least as well as — and with complementary signal to — existing CSF or PET biomarkers, contributing independent predictive variance beyond standard cognitive composites. | Prospective longitudinal cohort with repeat MSI testing, baseline and follow-up neuroimaging and CSF/PET biomarkers, and clinical outcome assessment. |
| P3 | The width of the audiovisual temporal binding window will be widened in individuals with preclinical AD (positive amyloid PET, cognitively unimpaired) relative to amyloid-negative controls of the same age and education, and will correlate with degree of amyloid burden. | Cross-sectional study in amyloid-stratified cohort using a validated smartphone-based TBW paradigm (simultaneity judgment or temporal order judgment task). |
| P4 | Digital MSI task performance (smartphone-based cross-modal reaction time, adaptation, and conflict paradigms) will correlate with laboratory MSI measurements at a level sufficient for clinical utility (r > 0.70 for continuous measures). | Validation study comparing digital MSI protocols to existing laboratory standards in the same participants. |
| P5 | Longitudinal tracking of digital MSI metrics over 12–24 months will show greater temporal sensitivity to early AD-related change than episodic memory tests of comparable burden administered at equivalent intervals. | Head-to-head longitudinal comparison of digital MSI metrics vs. validated brief cognitive instruments in an at-risk cohort. |
Digital Measurement Architecture
6.1 The Smartphone as MSI Assessment Platform
Laboratory MSI paradigms have historically required specialised audiovisual display equipment, calibrated audio hardware, and controlled experimental environments — conditions incompatible with large-scale population screening or continuous longitudinal monitoring. Modern smartphones offer a partial but meaningful solution. They provide three independently controllable sensory channels:
- Visual — high-refresh touchscreen, controllable luminance and chromatic contrast
- Auditory — stereo speaker and headphone output, ms-accurate audio playback via AudioTrack API
- Haptic / tactile — vibration actuator with precise timing control
- Touch response — capacitive touch with sub-frame timestamp resolution
- Display-to-photon latency varies across devices (~10–50 ms); calibration required
- Loudspeaker audio timing less precise than headphone; over-ear headphones preferred
- Vibration actuator latency less characterised than audio
- Home testing introduces noise variability not present in laboratory settings
- These limitations are manageable for relative tracking over time within an individual
6.2 Candidate Assessment Paradigms
The following paradigm types are compatible with smartphone delivery and map directly onto established laboratory MSI measures:
- Audiovisual simultaneity judgment (TBW measurement): A visual flash and an auditory tone are presented at variable stimulus onset asynchronies (SOAs). The user indicates whether they perceive the two stimuli as simultaneous or sequential. The resulting psychometric function yields the TBW width — the primary index of temporal binding.
- Cross-modal simple reaction time: Stimuli are delivered in unimodal (visual only, auditory only, haptic only) and bimodal (visual + auditory, visual + haptic) conditions. The multisensory reaction time benefit (bimodal advantage) quantifies integration efficiency.
- Audiovisual conflict task: A visual stimulus and an auditory stimulus are briefly presented at locations that are either spatially congruent or incongruent. Congruency effects on response time and error rate index the degree to which the brain automatically integrates cross-modal spatial information.
- Cross-modal adaptation and recalibration: After a period of exposure to systematically asynchronous audiovisual stimuli, the subsequent shift in the perceived point of subjective simultaneity measures temporal recalibration — an index of synaptic plasticity within MSI circuits that may be disproportionately sensitive to early cholinergic dysfunction in AD.
- Multisensory speech intelligibility: Spoken syllables are presented in noise at varying signal-to-noise ratios, with or without a matching lip-movement video. The audiovisual speech benefit in noise conditions measures real-world MSI utility under degraded conditions, directly relevant to social communication.
Absolute performance values on a smartphone will differ from laboratory measurements due to hardware variability. However, the relevant clinical signal for longitudinal monitoring is within-subject change over time, not cross-subject comparison at a single time point. Smartphone-based MSI assessment is therefore better suited for individual trajectory monitoring than for population-level threshold setting — a distinction that aligns naturally with the monitoring use case envisaged for NeuroFlex.
NeuroFlex as a Sensory Integration Monitoring Platform
NeuroFlex is designed as a daily engagement platform for individuals managing cognitive ageing and neurodegenerative risk. Its architecture — combining structured daily activities, routine-building features, cognitive challenges, and social engagement tracking — already captures many of the behavioural markers relevant to Functional Cognitive Reserve (see Paper I).
The SIRH identifies a distinct and complementary measurement opportunity: the integration of brief, low-burden multisensory assessment tasks into the daily NeuroFlex session, enabling continuous longitudinal tracking of MSI function alongside the cognitive and behavioural markers already collected.
Because MSI tasks are intrinsically brief (single trials require only 200–400 ms; a complete paradigm can be administered in under two minutes), they are suitable for integration into a daily engagement session without creating meaningful user burden. Their game-like properties — reaction to combined audiovisual or audio-haptic events — make them naturally compatible with an interactive daily engagement format.
Specifically, the following MSI measurement modalities could be implemented within the current NeuroFlex hardware and software architecture in a future research phase:
- Audiovisual simultaneity judgment (screen flash + tone burst)
- Cross-modal simple reaction time (visual + audio or visual + vibration)
- Auditory-haptic temporal order judgment
- Visual-haptic conflict response
- Audiovisual speech intelligibility (noise + lip video)
- Calibrated audio via in-app headphone compensation
- Olfactory assessment via structured caregiver-reported smell identification (questionnaire-based)
- Gait and balance via accelerometer (with consent)
- Eye tracking via front camera (device-dependent)
- Cross-modal adaptation and recalibration protocols
The research value of longitudinal MSI data from a large cohort of NeuroFlex users would be substantial, particularly if linked to periodic clinical assessments or biomarker data from willing participants in a future clinical research partnership. The combination of daily MSI tracking, behavioural engagement monitoring, and (in a research context) clinical outcome data would provide an unprecedented longitudinal dataset for validating the SIRH's predictions at population scale.
NeuroFlex Research Programme: A Coherent Line
The SIRH is designed to sit naturally within a broader, coherent research programme rather than appearing as an isolated proposition. The logical sequence of NeuroFlex's theoretical output follows an escalating logic: from the question of what protects against cognitive decline, to the question of what behaviours sustain that protection, to the question of what functional systems — beyond cognition itself — carry early diagnostic signal.
Discussion
The SIRH occupies a gap at the intersection of two established but largely separate research traditions: the literature on sensory changes in AD (primarily concerned with individual modalities) and the literature on multisensory integration (primarily developed in healthy ageing and developmental populations). Its contribution is to ask whether the integrative capacity itself — rather than any single input channel — carries early AD signal, and whether that signal is accessible via consumer digital hardware.
The clinical motivation for this approach is considerable. Existing early biomarkers for AD — amyloid and tau PET, CSF phospho-tau, plasma p-tau217 — are powerful but remain costly, invasive, or inaccessible outside specialist centres. Digital behavioural markers have attracted substantial recent interest as scalable, longitudinal complements to these modalities [27, 28]. The SIRH proposes a specific, mechanistically grounded target for digital assessment that is not reducible to standard cognitive testing.
The relationship between MSI and the auditory experience of early AD deserves particular emphasis. Many families of individuals with early AD describe a perception that their relative has "become hard of hearing" — when audiometric assessment reveals largely preserved peripheral thresholds. This observation is consistent with the hypothesis that central auditory processing and cross-modal integration are compromised, producing a functional experience of hearing difficulty that is not captured by standard audiometry. Digital tools capable of assessing speech-in-noise performance and audiovisual speech integration could therefore address a clinically visible phenomenon that current assessment tools systematically miss.
The SIRH also has implications for intervention. If MSI efficiency can be trained and maintained — as evidence from musical training and audiovisual speech training suggests [21] — then engagement with activities that require cross-modal coordination may represent a specific, mechanistically justified form of SIR maintenance. This aligns with the broader NeuroFlex philosophy of daily engagement as a functional protective measure, extending the Functional Reserve concept from Paper I into the sensory domain.
Limitations
10.1 Specificity of MSI Changes to AD
MSI efficiency declines in normal ageing, and is also affected by other neurological conditions. The SIRH does not claim that MSI deficits are specific to AD; rather, it claims that their rate of change, their correlation with AD-specific biomarkers, and their temporal precedence relative to cognitive decline are informative for AD risk monitoring. Establishing diagnostic specificity will require studies in cohorts including other dementia syndromes and non-AD cognitive impairment.
10.2 Device Variability
Smartphone hardware varies substantially across models and manufacturers in screen latency, audio output timing, and vibration precision. Any deployment of smartphone MSI assessment at scale will require either device-specific calibration or paradigm designs that are robust to these sources of variability. Per-device baseline measurement and within-subject tracking can partially mitigate this.
10.3 Confounds from Peripheral Sensory Loss
Age-related peripheral hearing and vision loss are highly prevalent in the populations most relevant to AD monitoring and will contaminate MSI measures if not controlled. Co-measurement of peripheral thresholds — via brief self-administered audiometric and visual acuity screening — is a necessary component of any MSI assessment protocol in older adults.
Conclusion
Alzheimer's disease changes not only what a person can remember, but how they perceive and integrate the world around them. The Sensory Integration Reserve Hypothesis proposes that this perceptual change — specifically, the degradation of the brain's capacity to bind concurrent signals from vision, hearing, and touch into coherent, temporally aligned percepts — begins before conventional cognitive assessment can detect it, and tracks with the underlying neuropathological burden of the disease.
If this hypothesis is correct, a family member who notices that their parent seems to have "become hard of hearing" may be observing not an audiological problem but an early signal of central neural change — a disruption not of the ear, but of the integrative machinery that makes heard speech comprehensible in a world full of competing sounds, competing movements, and competing sensory demands. That signal, properly measured and tracked over time, may carry diagnostic information that the current generation of cognitive tests cannot.
The smartphone in the user's pocket — capable of delivering precisely timed visual, auditory, and haptic stimuli and recording millisecond-accurate responses — is a sufficient platform to begin measuring that signal at scale. Building the measurement architecture, validating it against established biomarkers, and integrating it into a longitudinal engagement framework represents the next stage of the NeuroFlex research programme.
References
- Mendez MF, Ghajarania M, Perryman KM. Posterior cortical atrophy: clinical characteristics and differences compared to Alzheimer's disease. Dementia and Geriatric Cognitive Disorders. 2002;14(1):33–40.
- Geldmacher DS, Whitehouse PJ. Evaluation of dementia. New England Journal of Medicine. 1996;335(5):330–336.
- Stein BE, Meredith MA. The Merging of the Senses. Cambridge, MA: MIT Press, 1993.
- NeuroFlex Research Programme. The Dynamic Cognitive Reserve Hypothesis: Functional Maintenance, Not Lifetime Capacity, Determines Clinical Resilience in Alzheimer's Disease. NeuroFlex Research Series, Paper I, 2026. Available at: neuroflex.health/dynamic-cognitive-reserve/
- Doty RL. Olfactory dysfunction in Parkinson disease. Nature Reviews Neurology. 2012;8(6):329–339.
- Devanand DP, Michaels-Marston KS, Liu X, et al. Olfactory deficits in patients with mild cognitive impairment predict Alzheimer's disease at follow-up. American Journal of Psychiatry. 2000;157(9):1399–1405.
- Lin FR, Metter EJ, O'Brien RJ, et al. Hearing loss and incident dementia. Archives of Neurology. 2011;68(2):214–220.
- Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. The Lancet. 2020;396(10248):413–446.
- Rizzo M, Nawrot M. Perception of movement and shape in Alzheimer's disease. Brain. 1998;121(Pt 12):2259–2270.
- Hausdorff JM, Yogev G, Springer S, Simon ES, Giladi N. Walking is more like catching than tapping: gait in the elderly as a complex cognitive task. Experimental Brain Research. 2005;164(4):541–548.
- Laviola G, Hof PR, Pizzorusso T, et al. Somatosensory processing in neurodegenerative disorders. Neuroscience & Biobehavioral Reviews. 2017;77:380–391.
- Coughlan G, Laczó J, Hort J, Minihane AM, Hornberger M. Spatial navigation deficits — overlooked cognitive marker for preclinical Alzheimer disease? Nature Reviews Neurology. 2018;14(8):496–506.
- McGurk H, MacDonald J. Hearing lips and seeing voices. Nature. 1976;264(5588):746–748.
- Calvert GA, Spence C, Stein BE (eds). The Handbook of Multisensory Processes. Cambridge, MA: MIT Press, 2004.
- Whitwell JL, Przybelski SA, Weigand SD, et al. 3D maps from multiple MRI illustrate changing atrophy patterns as subjects progress from mild cognitive impairment to Alzheimer's disease. Brain. 2007;130(Pt 7):1777–1786.
- Meredith MA, Stein BE. Interactions among converging sensory inputs in the superior colliculus. Science. 1983;221(4608):389–391.
- Stevenson RA, Zemtsov RK, Wallace MT. Individual differences in the multisensory temporal binding window predict susceptibility to audiovisual illusions. Journal of Experimental Psychology: Human Perception and Performance. 2012;38(6):1517–1529.
- Setti A, Burke KE, Kenny RA, Newell FN. Is inefficient multisensory processing associated with falls in older people? Experimental Brain Research. 2011;209(3):375–384.
- Stern Y. Cognitive reserve in ageing and Alzheimer's disease. The Lancet Neurology. 2012;11(11):1006–1012.
- Stern Y, Barnes CA, Grady C, Jones RN, Raz N. Brain reserve, cognitive reserve, compensation, and maintenance: operationalization, validity, and mechanisms of cognitive resilience. Neurobiology of Aging. 2019;83:124–129.
- Strait DL, Kraus N. Can you hear me now? Musical training shapes functional brain networks for selective auditory attention and hearing speech in noise. Frontiers in Psychology. 2011;2:113.
- Kral A, Sharma A. Developmental neuroplasticity after cochlear implantation. Trends in Neurosciences. 2012;35(2):111–122.
- Brier MR, Thomas JB, Snyder AZ, et al. Loss of intranetwork and internetwork resting state functional connections with Alzheimer's disease progression. Journal of Neuroscience. 2012;32(26):8890–8899.
- Buckner RL, Andrews-Hanna JR, Schacter DL. The brain's default network: anatomy, function, and relevance to disease. Annals of the New York Academy of Sciences. 2008;1124:1–38.
- Klunk WE, Engler H, Nordberg A, et al. Imaging brain amyloid in Alzheimer's disease with Pittsburgh Compound-B. Annals of Neurology. 2004;55(3):306–319.
- Friston K. The free-energy principle: a unified brain theory? Nature Reviews Neuroscience. 2010;11(2):127–138.
- Doraiswamy PM, Sperling RA, Coleman RE, et al. Amyloid-β assessed by florbetapir F 18 PET and 18-month cognitive decline. Neurology. 2012;79(16):1636–1644.
- Kourtis LC, Regele OB, Wright JM, Jones GB. Digital biomarkers for Alzheimer's disease: the mobile/wearable devices opportunity. npj Digital Medicine. 2019;2(1):9.