Personalised Light Stimulation for Alzheimer’s Disease: Optimisation of Novel Solutions

Alzheimer’s disease is the most prevalent cause of dementia, and without a reliable therapy, it remains one of the most devastating and societally costly diseases. Decades of research has resulted in some breakthroughs with immunotherapy, which have shown a possibility for removing the characteristic amyloid plaques using monoclonal antibodies. However, their limited efficacy across multiple clinical trials of varying pharmaceutical substances and the disproportionately high incidence of side effects warrant continued investigations into alternative or supplementary therapies.

Forty hertz sensory gamma stimulation may offer a promising approach, as recent evidence suggests its ability to modulate glymphatic clearance of amyloid. However, exploiting sensory pathways for non-invasive intervention brings an inherent challenge, as stimulation can be more or less comfortable on the senses. In visual stimulation schemes, the exposure to bright and flashing lights is associated with biological side effects, driven in part by the sensation of flicker. This challenge may ultimately result in a poorer degree of adherence to therapy, rendering it less effective. The World Health Organisation estimates that 50% of patients do not adhere to chronic medication and identify a need to tailor therapy to the individual. This motivates a search for visual stimuli which reduce the flickering sensation while maintaining the ability to evoke a cortical gamma response. Given that people perceive sensations differently, personalised stimulation could further this goal.

Despite an unclear understanding for the mechanism of action and incomplete clinical evidence for its efficacy, it is hypothesised that improvements can be made to the first iteration(s) of 40 Hz visual stimulation therapy.

This thesis seeks to investigate how a personally optimised visual stimulus might be reached. There are many challenges associated with optimising a potential therapy. Here, the possibility of using short-term surrogate neurophysiological optimisation objectives in place of clinical endpoints – and the assumptions needed to do so – are discussed. Likewise, the vast search space of configurable stimulus parameters is explored, and a subset of configurations are tested in a range of experiments to constrain the solution. In addition, the recipient of the stimulus is taken into account as an agent whose behaviour may affect stimulation.

The sanctity of the 40 Hz stimulation frequency is challenged by its historical origins, and, in a experiment comparing evoked cortical responses to stimulation across the 36 to 44 Hz band, no differences are found. For reducing flicker perception, it is advantageous to increase the frequency, but it remains uncertain what the clinical implications might be, rendering interpretation of the results difficult.

When comparing 40 Hz heterochromatic flicker with various colour combinations, the extremes of the visible spectrum are associated with a stronger cortical response than the mid-spectrum colours. It is concluded that including either red and/or blue is more important than using spectrally dissimilar colours.

In a comparison of three distinct types of 40 Hz flicker, namely heterochromatic, luminance, and invisible spectral flicker, they scored differently on both comfort and perceived flicker. The power of their 40 Hz cortical responses are proportional to both, thereby presenting a trade-off between user experience and brain response. Importantly, this association does not exist within each type of flicker, in which the power varies independently from comfort and perceived flicker.

To free the user from needing to look at the stimulation device, it is confirmed that indirect 40 Hz stimulation at angles of up to 30◦ evokes significant cortical responses. This offers flexibility in therapy, and the observed decline in response with increasing angle may be automatically compensated with additional time based on in-device gaze tracking. Another experiment aimed to validate and build on evidence showing that cognitively demanding tasks during stimulation modulates the 40 Hz response. It failed due to electromagnetic incompatibility, and instead, the technical details and suggestions for future replication are presented.

While a closed-loop stimulus optimisation scheme has yet to be realised, the results present guidance for parameterising and constraining the solutions. To achieve this, an implementation based on electroencephalography and multi-objective Bayesian optimisation is suggested to account for noisy observations in cortical response and comfort. The era of digital therapeutics may lead to further development of stimulation and – given successful clinical trials – allow for continuous improvement of therapeutic devices for Alzheimer’s disease.

Future research into the biological understanding for why and how gamma stimulation affects Alzheimer’s disease may provide further clarity on which optimisation objectives are important. In conjunction with collective reporting on clinical- and neurophysiological endpoints, the uncertainty in predictions of efficacy from short-term surrogates will be better understood.

Ultimately, the aspects investigated in this thesis have provided several insights into factors that may affect the effectiveness of gamma stimulation and tolerability of therapy. With due diligence for understanding those factors, controllable by either device or user, the future of personalized therapy is one step closer. However, closed-loop neuroimaging based optimisation of visual gamma stimulation paradigms still requires further maturation on several levels. This includes its assumptions, parametrisation, and noise levels, rendering the technique promising but not yet realised.