Mild cognitive impairment, dementia and Parkinson's disease
Neurodegenerative conditions are not in the AAPB 2023 book, so every level shown here is NeuroLogic's own. The picture splits in two. For mild cognitive impairment and early dementia, EEG neurofeedback shows moderate memory gains across a dozen small trials, and one larger trial combining cognitive training with fNIRS feedback did better than a sham condition. For Parkinson's disease, people clearly learn to change their brain rhythms, but this has not yet translated into reliable improvement of motor symptoms; muscle or postural biofeedback added to physiotherapy gives at best a small extra gain in balance. Heart-rate-variability breathing is an early but interesting lead for cognition and Alzheimer-related blood markers.
Updated :
What the research shows
EEG neurofeedback — standard amplitude training
(alpha, SMR/theta, gamma protocols) — mild cognitive impairment and dementia
Ages
Adult (18+)
Techniques
Alpha training, Sensorimotor rhythm (SMR), Other neurofeedback
AAPB0Not ratedNeuroLogic3Probably efficacious
Level 3. Lin 2024 pooled 14 clinical trials (n = 284) of EEG neurofeedback in healthy older adults and people with mild cognitive impairment (MCI): working memory Hedges' g = 0.665 (95 % CI 0.473 to 0.858) and episodic memory g = 0.595 (0.333 to 0.856), with the MCI subgroup significant on both (working memory g = 0.812, k = 6; episodic memory g = 0.503, k = 6) and benefit confined to programmes totalling more than 300 minutes of training. Tazaki 2024 (13 studies in MCI, Alzheimer's and vascular cognitive impairment) reports gains in memory or attention in every study, without formal quality appraisal. Tsai 2025 (16 EEG-BCI/neurofeedback trials in older adults with or without MCI) finds the positive results biased by randomisation and outcome measurement. A double-blind sham-controlled study in older adults with subjective memory complaints (Andrade 2024) shows protocol-specific self-modulation of gamma synchronisation without cognitive change. Multiple controlled studies converge on a moderate cognitive benefit, but the trials are small, heterogeneous in protocol and site, and no adequately powered randomised trial has been replicated in an independent setting — hence Level 3 rather than 4.
Sensorimotor rhythm (SMR), Beta training, Alpha training
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. Von Altdorf 2025, a pre-registered systematic review and meta-analysis of 11 EEG neurofeedback studies in Parkinson's disease (143 participants, Hoehn and Yahr I-IV), finds that cortical activity changes in the prescribed direction (SMD 1.30, 95 % CI 0.50 to 2.10, high certainty) but that the effect on motor symptoms is inconclusive (SMD 0.10, 95 % CI −1.03 to 1.23, low certainty). Romero 2024, a single-blind four-arm randomised trial (n = 40; eight 30-minute sessions reducing bilateral alpha and beta, rTMS, both combined, or no intervention), reports the largest UPDRS-III and quality-of-life effects in the combined arm, followed by rTMS alone and neurofeedback alone, with negligible differences from the untreated arm for functional mobility and postural stability. Mirabella 2026 (12 EEG, fMRI and deep-brain-stimulation-signal neurofeedback studies retained) concludes that patients learn to modulate the targeted signals but that translation to validated clinical outcomes is limited and inconsistent. Target engagement is established and a small randomised trial points in the right direction; clinical efficacy is not.
Level 2. Tinaz 2022 randomised 44 non-demented people with mild Parkinson's disease (Hoehn and Yahr ≤ 3) to fMRI-neurofeedback-guided kinesthetic motor imagery (10-12 runs targeting right insula–dorsomedial frontal connectivity, plus four weeks of daily home imagery) or visual imagery: no significant difference on the primary imaging outcome or on the MDS-UPDRS motor exam, comparable improvement in motor function scores in both arms (secondary outcome), and connectivity-based regulation judged unsuccessful. An earlier small randomised trial from the same group, Subramanian 2016, reported motor gains after supplementary-motor-area neurofeedback [to verify]. Mirabella 2026 counts four fMRI studies among the 12 retained and finds self-regulation achievable but clinical translation limited. Feasibility is shown; efficacy is not, and no independent replication exists.
Level 2. Park 2025 randomised 90 older adults with mild cognitive impairment to virtual-reality cognitive training delivered in parallel with fNIRS-derived prefrontal neurofeedback, a sham condition, or a waitlist (eight 15-minute sessions): the active group improved more than both controls on the Trail Making Test B and backward digit span and showed lower dorsolateral prefrontal activity during testing, read as greater neural efficiency; effect sizes are not reported in the abstract. Lee 2023 (n = 13, four weekly sessions of cognitive training with left-DLPFC fNIRS neurofeedback, no control) reports working-memory gains correlating with post-training oxygenation change. One sham-superior randomised trial from a single group, in which neurofeedback is inseparable from the cognitive-training package and the dose is very short, supports Level 2; independent replication is needed for more.
Biofeedback
EMG, postural and other biofeedback — Parkinson's disease (motor and non-motor symptoms)
Ages
Adult (18+)
Techniques
EMG biofeedback, Other biofeedback
AAPB0Not ratedNeuroLogic3Probably efficacious
Level 3. Baumgart 2026 pooled 20 randomised trials (814 participants) in which biofeedback was added to practice of sitting, standing or walking against the same practice alone: the Berg Balance Scale improved by 1.5 points out of 56 (95 % CI 0.7 to 2.3, high-quality evidence), while Timed Up and Go (−0.2 s, 95 % CI −1.4 to 1.8) and walking speed (0.02 m/s, 95 % CI −0.06 to 0.1) did not differ; the authors conclude that biofeedback is no more effective than practice alone. Yakşi 2022 (n = 41, posturography-assisted biofeedback plus conventional balance exercises versus exercises alone, six weeks) found no significant between-group difference on any measure. Non-motor targets are more encouraging but thinly studied: Roßkopf 2024 (quasi-randomised, 34 in-patients) reports greater gains in facial expression and emotion recognition with facial EMG biofeedback than with non-facial exercises; Chen 2025 (waitlist-controlled pilot, n = 19, modality unspecified) reports reduced anxiety persisting at one month. Gandhi 2022 rates the evidence for exercises with biofeedback in dysphagia as low certainty. Many randomised trials, a statistically significant but clinically marginal balance effect, and no consistent superiority over practice alone: Level 3.
HRV and slow-breathing biofeedback — mild neurocognitive disorder and Alzheimer's biomarkers
Level 2. Manser 2024 (Brain-IT) randomised people with mild neurocognitive disorder to usual care with or without a co-designed exergame motor-cognitive programme that includes resonance breathing guided by HRV biofeedback: large, statistically significant effects on global cognition and immediate and delayed verbal recall, with 55 % of participants showing a clinically relevant improvement; the sample size is not given in the abstract and HRV biofeedback is one component of a multimodal package. In healthy adults, Min 2023 (n = 108, four weeks of daily HRV biofeedback to increase versus decrease heart-rate oscillations) produced large between-condition differences in plasma amyloid-beta 40 and 42, partly replicated by Nashiro 2026 (n = 62, aged 50-70, ten weeks of slow-paced versus random-paced breathing: greater decrease in Aβ42 only, no effect on brain structure or cognition). Allen 2026 (53 healthy older adults, five weeks, sham-controlled) found no specific gain in cognition or emotional functioning. One randomised trial in a clinical sample with the feedback embedded in a package, plus biomarker-only trials in healthy volunteers: Level 2.
By age
Adult (18+)
All the evidence is in adults, mostly over 60; there is no paediatric literature and the levels apply to adults only.
In short
Clinical reading
NeuroLogic Level 3 for amplitude neurofeedback in MCI and dementia (Lin 2024 meta-analysis: working memory g = 0.67, episodic memory g = 0.60, MCI subgroup significant, benefit above 300 min of training), Level 2 in Parkinson's disease (von Altdorf 2025: cortical modulation SMD 1.30, motor symptoms SMD 0.10, inconclusive). fMRI (Tinaz 2022) and fNIRS (Park 2025) neurofeedback: Level 2. Biofeedback: Level 3 for EMG/postural biofeedback in Parkinson's motor rehabilitation (Baumgart 2026: 20 RCTs, balance +1.5/56, mobility and gait null), Level 2 for HRV biofeedback in mild neurocognitive disorder (Manser 2024).
Protocols
MCI/dementia: alpha, SMR/theta or gamma protocols, often paired with cognitive training; benefit appears only above 300 cumulative minutes (≈ 20-30 sessions). Parkinson's: SMR up or beta/alpha reduction over sensorimotor cortex; fMRI targets the supplementary motor area or insula-frontal connectivity. Biofeedback: EMG or force-plate feedback during balance and gait practice, 6-8 weeks; resonance-frequency breathing with daily home practice.
Limits
No AAPB rating; small heterogeneous trials with unclear blinding; MCI trials pool healthy and impaired older adults; in Parkinson's, neural change without clinical change; biofeedback in Parkinson's motor rehabilitation adds little to matched practice; HRV findings rest on biomarkers in healthy volunteers or on a multimodal package; no paediatric evidence (adult-only indication); ALS and other neurodegenerative diseases have no neurofeedback or biofeedback trial in the 2022-2026 window.
Study base
Not in the AAPB 2023 book. 2022-2026 base: 14 publications indexed in the archive (10 neurofeedback, 4 biofeedback/HRV), including three meta-analyses or systematic reviews in Parkinson's disease and one meta-analysis in MCI.
Brendan's perspective
No AAPB level, so both numbers are ours, and they should not be the same. For mild cognitive impairment and dementia, Lin 2024 supports a defensible 3: the MCI subgroup is significant on both memory outcomes, again only above 300 minutes of training. Tsai 2025 is right that randomisation and outcome measurement carry the bias, so I would not go higher. Parkinson's is a 2 and deserves it. Von Altdorf 2025 finds cortical activity moving in the prescribed direction with high certainty (SMD 1.30) and motor symptoms not moving at all (SMD 0.10) — target engagement without clinical benefit, which is why I keep the trained metric and the outcome in separate columns. Romero 2024 ran eight sessions, which is not a course. Baumgart 2026 is the sober note on the biofeedback side: 1.5 balance points out of 56, no better than the same practice without feedback. In early cognitive complaint I would train, qEEG-based, and say plainly that this is function and compensation, not a modified disease trajectory. This is an area that badly needs more research, and the encouraging thing is that the trend of the last few years is running that way. My own expectation, stated so it can be tested rather than believed: neurofeedback can contribute to maintaining cognitive capacity, and it can bolster emotional self-regulation, which matters enormously to how these conditions are lived. What I do not think it can do is reverse losses already sustained. Stabilising them looks plausible; eking out some gain through compensatory routes looks plausible; recovering what has gone does not. And there is a question underneath all of it that nobody has designed a study to answer: are we postponing cognitive decline, or damping its slope while the endpoint arrives on schedule? Those two look identical at twelve months and completely different at five years, and until somebody follows a cohort long enough to tell them apart, anyone claiming either is guessing.
Lin et al. (2024) Effectiveness of Electroencephalography Neurofeedback for Improving Working Memory and Episodic Memory in the Elderly: A Meta-Analysis doi:10.3390/medicina60030369
von Altdorf et al. (2025) Effectiveness of Electroencephalographic Neurofeedback for Parkinson's Disease: A Systematic Review and Meta-Analysis doi:10.3390/jcm14196929
Romero et al. (2024) Clinical and neurophysiological effects of bilateral repetitive transcranial magnetic stimulation and EEG-guided neurofeedback in Parkinson's disease: a randomized, four-arm controlled trial doi:10.1186/s12984-024-01427-5
Tinaz et al. (2022) Neurofeedback-guided kinesthetic motor imagery training in Parkinson's disease: Randomized trial doi:10.1016/j.nicl.2022.102980
Park et al. (2025) Is virtual reality-based cognitive training in parallel with functional near-infrared spectroscopy-derived neurofeedback beneficial to improve cognitive function in older adults with mild cognitive impairment? doi:10.1080/09638288.2024.2380483
Baumgart et al. (2026) Biofeedback in Parkinson's disease: A systematic review with meta-analysis doi:10.1177/02692155261474067
Manser et al. (2024) "Brain-IT": Exergame training with biofeedback breathing in neurocognitive disorders doi:10.1002/alz.13913
Diotaiuti et al. (2025) Biofeedback for Motor and Cognitive Rehabilitation in Parkinson's Disease: A Comprehensive Review of Non-Invasive Interventions doi:10.3390/brainsci15070720
