Machine Learning Techniques for Developing Remotely Monitored Central Nervous System Biomarkers Using Wearable Sensors: A Narrative Literature Review.

BACKGROUND: Central nervous system (CNS) disorders benefit from ongoing monitoring to assess disease progression and treatment efficacy. Mobile health (mHealth) technologies offer a means for the remote and continuous symptom monitoring of patients. Machine Learning (ML) techniques can process and engineer mHealth data into a precise and multidimensional biomarker of disease activity. OBJECTIVE: This narrative literature review aims to provide an overview of the current landscape of biomarker development using mHealth technologies and ML. Additionally, it proposes recommendations to ensure the accuracy, reliability, and interpretability of these biomarkers. METHODS: This review extracted relevant publications from databases such as PubMed, IEEE, and CTTI. The ML methods employed across the selected publications were then extracted, aggregated, and reviewed. RESULTS: This review synthesized and presented the diverse approaches of 66 publications that address creating mHealth-based biomarkers using ML. The reviewed publications provide a foundation for effective biomarker development and offer recommendations for creating representative, reproducible, and interpretable biomarkers for future clinical trials. CONCLUSION: mHealth-based and ML-derived biomarkers have great potential for the remote monitoring of CNS disorders. However, further research and standardization of study designs are needed to advance this field. With continued innovation, mHealth-based biomarkers hold promise for improving the monitoring of CNS disorders.

Transcranial magnetic stimulation as biomarker of excitability in drug development: A randomized, double-blind, placebo-controlled, cross-over study.

AIMS: The purpose of this study was to investigate pharmacodynamic effects of drugs targeting cortical excitability using transcranial magnetic stimulation (TMS) combined with electromyography (EMG) and electroencephalography (EEG) in healthy subjects, to further develop TMS outcomes as biomarkers for proof-of-mechanism in early-phase clinical drug development. Antiepileptic drugs presumably modulate cortical excitability. Therefore, we studied effects of levetiracetam, valproic acid and lorazepam on cortical excitability in a double-blind, placebo-controlled, 4-way cross-over study. METHODS: In 16 healthy male subjects, single- and paired-pulse TMS-EMG-EEG measurements were performed predose and 1.5, 7 and 24 hours postdose. Treatment effects on motor-evoked potential, short and long intracortical inhibition and TMS-evoked potential amplitudes, were analysed using a mixed model ANCOVA and cluster-based permutation analysis. RESULTS: We show that motor-evoked potential amplitudes decreased after administration of levetiracetam (estimated difference [ED] -378.4 µV; 95%CI: -644.3, -112.5 µV; P < .01), valproic acid (ED -268.8 µV; 95%CI: -532.9, -4.6 µV; P = .047) and lorazepam (ED -330.7 µV; 95%CI: -595.6, -65.8 µV; P = .02) when compared with placebo. Long intracortical inhibition was enhanced by levetiracetam (ED -60.3%; 95%CI: -87.1%, -33.5%; P < .001) and lorazepam (ED -68.2%; 95%CI: -94.7%, -41.7%; P < .001) at a 50-ms interstimulus interval. Levetiracetam increased TMS-evoked potential component N45 (P = .004) in a central cluster and decreased N100 (P < .001) in a contralateral cluster. CONCLUSION: This study shows that levetiracetam, valproic acid and lorazepam decrease cortical excitability, which can be detected using TMS-EMG-EEG in healthy subjects. These findings provide support for the use of TMS excitability measures as biomarkers to demonstrate pharmacodynamic effects of drugs that influence cortical excitability.

Transcranial magnetic stimulation as a translational biomarker for AMPA receptor modulation.

TAK-653 is a novel α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-positive allosteric modulator being developed as a potential therapeutic for major depressive disorder (MDD). Currently, there are no translational biomarkers that evaluate physiological responses to the activation of glutamatergic brain circuits available. Here, we tested whether noninvasive neurostimulation, specifically single-pulse or paired-pulse motor cortex transcranial magnetic stimulation (spTMS and ppTMS, respectively), coupled with measures of evoked motor response captures the pharmacodynamic effects of TAK-653 in rats and healthy humans. In the rat study, five escalating TAK-653 doses (0.1-50 mg/kg) or vehicle were administered to 31 adult male rats, while measures of cortical excitability were obtained by spTMS coupled with mechanomyography. Twenty additional rats were used to measure brain and plasma TAK-653 concentrations. The human study was conducted in 24 healthy volunteers (23 males, 1 female) to assess the impact on cortical excitability of 0.5 and 6 mg TAK-653 compared with placebo, measured by spTMS and ppTMS coupled with electromyography in a double-blind crossover design. Plasma TAK-653 levels were also measured. TAK-653 increased both the mechanomyographic response to spTMS in rats and the amplitude of motor-evoked potentials in humans at doses yielding similar plasma concentrations. TAK-653 did not affect resting motor threshold or paired-pulse responses in humans. This is the first report of a translational functional biomarker for AMPA receptor potentiation and indicates that TMS may be a useful translational platform to assess the pharmacodynamic profile of glutamate receptor modulators.

Spatiotemporal Dynamics of Single and Paired Pulse TMS-EEG Responses.

For physiological brain function a particular balance between excitation and inhibition is essential. Paired pulse transcranial magnetic stimulation (TMS) can estimate cortical excitability and the relative contribution of inhibitory and excitatory networks. Combining TMS with electroencephalography (EEG) enables additional assessment of the spatiotemporal dynamics of neuronal responses in the stimulated brain. This study aims to evaluate the spatiotemporal dynamics and stability of single and paired pulse TMS-EEG responses, and assess long intracortical inhibition (LICI) at the cortical level. Twenty-five healthy subjects were studied twice, approximately one week apart. Manual coil positioning was applied in sixteen subjects and robot-guided positioning in nine. Both motor cortices were stimulated with 50 single pulses and 50 paired pulses at each of the five interstimulus intervals (ISIs): 100, 150, 200, 250 and 300 ms. To assess stability and LICI, the intraclass correlation coefficient and cluster-based permutation analysis were used. We found great resemblance in the topographical distribution of the characteristic TMS-EEG components for single and paired pulse TMS. Stimulation of the dominant and non-dominant hemisphere resulted in a mirrored spatiotemporal dynamics. No significant effect on the TMS-EEG responses was found for either stimulated hemisphere, time or coil positioning method, indicating the stability of both single and paired pulse TMS-EEG responses. For all ISIs, LICI was characterized by significant suppression of the late N100 and P180 components in the central areas, without affecting the early P30, N45 and P60 components. These observations in healthy subjects can serve as reference values for future neuropsychiatric and pharmacological studies.

Infraslow activity as a potential modulator of corticomotor excitability.

Fluctuations in cortical excitability are a candidate mechanism involved in the trial-to-trial variation of motor evoked potentials (MEPs) to transcranial magnetic stimulation (TMS). We explore whether infraslow EEG activity (<0.1 Hz) modulates corticomotor excitability by evaluating the presence of temporal and phase clustering of TMS-induced MEPs. In addition, we evaluate the dependence of MEP amplitude on the phase of the infraslow activity. Twenty-three subjects were stimulated at an intensity above the resting motor threshold (rMT) and ten at the rMT. We evaluated whether temporal and phase clustering of MEP size and MEP generation were present, using 1,000 surrogates with a similar amplitude or occurrence distribution. To evaluate the MEP amplitude dependence, we used the least-square method to approximate the linear circular data by fitting a sine function. We observed significant temporal clustering at a group level, in all individual subjects stimulated at rMT and in the majority of those stimulated above rMT, suggesting underlying determinism of corticomotor excitability instead of randomly generated fluctuations. The majority of subjects showed significant phase clustering for MEP size and for MEP occurrence, and significant phase clustering was found at the group level. Furthermore, in approximately one-quarter to one-half of the subjects we found a significant correlation and dependence of MEP amplitude on the phase of infraslow activity, respectively. Although other mechanisms very likely contribute as well, our findings seem to suggest that infraslow activity is involved in the variability of cortical excitability and TMS-induced responses. NEW & NOTEWORTHY Cortical excitability measures are highly variable during transcranial magnetic stimulation. Although ongoing brain oscillations are assumed to modulate excitability, no consistent associations are found for the traditional frequency bands. We focus on the role of infraslow EEG activity, defined as rhythms with frequencies < 0.1 Hz. We provide experimental evidence suggesting that infraslow activity most likely modulates corticomotor excitability and that response variation could be reduced when stimulation is targeted at a specific infraslow phase.

Resting Motor Threshold, MEP and TEP Variability During Daytime.

Humans show a variation in physiological processes during the day. To reliably assess (changes in) cortical excitability with transcranial magnetic stimulation (TMS), it is relevant to know the natural variation in TMS readouts during the day. In case of significant daytime variations, this should be taken into account when scheduling (follow-up) measurements. This study aims to evaluate the influence of the time of day on the resting motor threshold (RMT), motor evoked potential (MEP) and TMS evoked potential (TEP) in healthy controls. TMS-EMG-EEG was recorded in 16 healthy subjects. At both motor cortices, we administered 75 pulses at an intensity of 110% RMT. Subjects were stimulated during five sessions in one day (8:00 AM, 10:30 AM, 1:00 PM, 3:30 PM and 6:00 PM) while keeping the stimulation intensity constant. We compared the TEP waveforms between the five sessions with a cluster-based permutation analysis, and the RMT and MEP amplitude with rmANOVA. In general there were no significant differences between the five sessions in the RMT, MEP amplitude or TEP. Only for the left side, N100 amplitude was larger at 3:30 PM than 10:30 AM. The standard deviation of the P30 and N100 amplitude was significantly higher between subjects within one session than within single subjects during the day. The TEP is highly reproducible during the day, with a low intra-individual variation compared to the inter-individual variation. In addition, we found no significant variation of the RMT and MEP amplitude between multiple sessions on one day.

Accurate Coil Positioning is Important for Single and Paired Pulse TMS on the Subject Level.

Function-guided navigation is commonly used when assessing cortical excitability using transcranial magnetic stimulation (TMS). However, the required accuracy, stability and the effect of a change in coil positioning are not entirely known. This study investigates the accuracy of function-guided navigation for determining the hotspot. Furthermore, it evaluates the effect of a change in coil location on the single and paired pulse excitability measures: motor evoked potential (MEP) amplitude, TMS evoked potential (TEP) and long intracortical inhibition (LICI), and of a change in coil orientation on LICI. Eight healthy subjects participated in the single pulse study, and ten in the paired pulse study. A robot-guided navigation system was used to ensure accurate and stable coil positioning at the motor hotspot as determined using function-guided navigation. In addition, we targeted four locations at 2 mm and four at 5 mm distance around the initially defined hotspot, and we increased and decreased the coil orientation by 10°. In none of the subjects, the largest MEP amplitudes were evoked at the originally determined hotspot, resulting in a poor accuracy of function-guided navigation. At the group level, a change in coil location had no significant effect on the MEP amplitude, TEP, or LICI, and a change in coil orientation did not significantly affected LICI. However, at the subject level significant effects on MEP amplitude, TEP, and LICI were found for changes in coil location or orientation, although absolute differences were relatively small and did not show a consistent pattern. This study indicates that a high accuracy in coil positioning is especially required to measure cortical excitability reliably in individual subjects using single or paired pulse TMS.

Long-interval intracortical inhibition as biomarker for epilepsy: a transcranial magnetic stimulation study.

Cortical excitability, as measured by transcranial magnetic stimulation combined with electromyography, is a potential biomarker for the diagnosis and follow-up of epilepsy. We report on long-interval intracortical inhibition data measured in four different centres in healthy controls (n = 95), subjects with refractory genetic generalized epilepsy (n = 40) and with refractory focal epilepsy (n = 69). Long-interval intracortical inhibition was measured by applying two supra-threshold stimuli with an interstimulus interval of 50, 100, 150, 200 and 250 ms and calculating the ratio between the response to the second (test stimulus) and to the first (conditioning stimulus). In all subjects, the median response ratio showed inhibition at all interstimulus intervals. Using a mixed linear-effects model, we compared the long-interval intracortical inhibition response ratios between the different subject types. We conducted two analyses; one including data from the four centres and one excluding data from Centre 2, as the methods in this centre differed from the others. In the first analysis, we found no differences in long-interval intracortical inhibition between the different subject types. In all subjects, the response ratios at interstimulus intervals 100 and 150 ms showed significantly more inhibition than the response ratios at 50, 200 and 250 ms. Our second analysis showed a significant interaction between interstimulus interval and subject type (P = 0.0003). Post hoc testing showed significant differences between controls and refractory focal epilepsy at interstimulus intervals of 100 ms (P = 0.02) and 200 ms (P = 0.04). There were no significant differences between controls and refractory generalized epilepsy groups or between the refractory generalized and focal epilepsy groups. Our results do not support the body of previous work that suggests that long-interval intracortical inhibition is significantly reduced in refractory focal and genetic generalized epilepsy. Results from the second analysis are even in sharper contrast with previous work, showing inhibition in refractory focal epilepsy at 200 ms instead of facilitation previously reported. Methodological differences, especially shorter intervals between the pulse pairs, may have contributed to our inability to reproduce previous findings. Based on our results, we suggest that long-interval intracortical inhibition as measured by transcranial magnetic stimulation and electromyography is unlikely to have clinical use as a biomarker of epilepsy.

Fluid Resuscitation in Septic Patients Improves Systolic but not Diastolic Middle Cerebral Artery Flow Velocity.

To investigate the effects of fluid resuscitation on cerebral hemodynamics in sepsis, the following set of transcranial Doppler (TCD) parameters was used: maximal change in flow velocity (FV) during stroke onset (acc), maximal FV during first (sys1) or second (sys2) phase of systole and mean diastolic FV (dias@560). We aim to evaluate changes in cerebral hemodynamics that result from (i) sepsis and (ii) adequate fluid resuscitation in critically ill septic patients. In the majority of 16 septic patients sys2 was initially absent but reappeared during the period of fluid resuscitation; whereas sys2 absence was never seen in healthy controls. Second, adequate fluid resuscitation resulted in a significant increase of the systolic FV components (acc, sys1, sys2 and systolic blood pressure); whereas the diastolic components (dias@560 and diastolic blood pressure) remained unchanged. Sys2 absence and reappearance in sepsis suggests that TCD could become a non-invasive alternative for hemodynamic monitoring.

Repeatability of long intracortical inhibition in healthy subjects.

OBJECTIVES: Transcranial magnetic stimulation (TMS) is widely used to assess cortical excitability. To detect changes in excitability with longitudinal studies, it is important to validate the repeatability of excitability measures within a subject between different sessions. Repeatability studies on long intracortical inhibition (LICI) are limited and reported agreement ranges from poor to good. This study aims to evaluate the repeatability of LICI in healthy subjects using paired pulse TMS. In addition, it investigates whether LICI repeatability differs for manual and robot-guided coil positioning. METHODS: Thirty healthy subjects (10 males, mean age 28.4 ± 8.2 years) were studied twice, approximately one week apart. Both motor cortices were stimulated with 50 paired pulses (intensity 120% of resting motor threshold) at interstimulus intervals (ISIs): 50, 100, 150, 200, 250 and 300 ms. In twenty subjects a figure-of-eight coil was positioned and held in place manually during both sessions, while in ten subjects a robot-navigated arm was used. LICI repeatability was assessed using the intraclass correlation coefficient (ICC). RESULTS: For manual and robot-guided coil positioning we found a large variation in repeatability at the subject level and ISI level, ranging from poor to good agreement. On a group level, we found good repeatability for averaged LICI curves (manual: ICC = 0.91, robot-guided: ICC = 0.95), which decreased when individual curves were correlated between sessions (manual: ICC = 0.76, robot-guided: ICC = 0.84). CONCLUSION: For a correct interpretation of longitudinal study outcomes it is important to know the subject specific LICI repeatability and to analyze each ISI individually. Furthermore, the added value of robot-guided coil positioning for paired pulse TMS seems limited. SIGNIFICANCE: The large variation in LICI repeatability at the subject level and ISI level should be taken into account in longitudinal studies, while robot-guided coil positioning seems unnecessary.

Single and paired pulse transcranial magnetic stimulation in drug naïve epilepsy.

Transcranial magnetic stimulation (TMS) measures cortical excitability and is therefore potentially suitable as an additional tool for epilepsy diagnostics and therapy evaluation. In this review we discuss the application of TMS in epilepsy research and systematically analyze single and paired pulse TMS outcomes from 31 drug naïve patient studies. Despite a large variety in used TMS protocols, there was no relation between specific protocol aspects and the occurrence of significant results. Protocols were often not in accordance with latest guidelines and recommendations. Cortical excitability, as measured by TMS, was increased in drug naïve epilepsy patients, being most prominent for generalized epilepsy. Single pulse TMS indicated a trend towards a lower resting motor threshold (rMT) and a prolonged cortical silent period (CSP) for generalized epilepsy, while inconclusive results were found for focal epilepsy. The paired pulse TMS outcomes, short intracortical inhibition (SICI) and long intracortical inhibition (LICI), showed the most consistent significant increase in cortical excitability in generalized and focal epilepsy patients. Future epilepsy research should especially focus on the interstimulus intervals 2 and 5ms for SICI, and 250 and 300ms for LICI. Furthermore, combining TMS with electroencephalography (EEG) may contribute to analysis on an individual patient level.