Stimulus-induced Gamma rhythms are weaker in human elderly with Mild Cognitive Impairment and Alzheimer's Disease

Alzheimer's Disease (AD) in elderly adds substantially to socio-economic burden necessitating early diagnosis. While recent studies in rodent models of AD have suggested diagnostic and therapeutic value for gamma rhythms in brain, the same has not been rigorously tested in humans. We recruited a large population (N=247; 106 females) of elderly (>49 years) individuals from the community, who viewed large gratings that induced strong gamma oscillations in their electroencephalogram (EEG). These individuals were classified as healthy (N=227), mild-cognitively-impaired (MCI; 14) or AD (6) based on clinical history and Clinical Dementia Rating scores. Surprisingly, stimulus-induced gamma rhythms, but not alpha or steady-state-visually-evoked-responses, were significantly lower in both MCI and AD patients compared to their age and gender matched controls. This reduction was not due to differences in eye movements or baseline power. Our results suggest that gamma could be used as potential diagnostic tool for MCI/AD in humans.


Introduction 35
Alzheimer's Disease (AD) is a predominant cause of dementia (decline in cognitive 36 abilities) of old age and substantially contributes to the socio-economic burden in the geriatric 37 population, necessitating early diagnosis. Advances in our understanding of cellular pathology 38 of AD in rodent models and its link to gamma rhythms in brain has spurred interest to 39 investigate diagnostic and therapeutic potential of gamma rhythms in AD and other forms of

76
First, we examined how the two gamma rhythms differed in cases as compared to their 77 healthy age and gender matched controls using full-screen static sinusoidal grating stimuli 78 presented on a computer monitor. Out of the 227 cognitively healthy participants, we selected 79 114 controls who were age (± 1 year) and gender matched with their respective cases 80 (MCI/AD) for analysis. One AD patient (A1 in Supplementary Figure 1, 92 years old male) 81 did not have any matching control. 82 We averaged spectral data for all analyzable bipolar electrodes (as described in 17) 83 from 10 occipital and parieto-occipital pairs (marked in black enclosures in Figure 1d; see SI 84 Methods). Figure 1a shows the median stimulus-induced change in PSDs for 20 cases and 114  Figure 1a). This could also be seen in the median time-frequency change in power 91 spectrograms (baseline: -500-0 ms of stimulus onset) for cases and controls in Figure 1b. 92 Change in band-limited power was lesser for both gamma bands in the case group 93 compared to the control group ( Figure 1c, KW test, χ 2 (133)=6.14, p=0.013 for slow and 94 χ 2 (133)=8.00, p=0.005 for fast gamma, both significant at a Bonferroni-adjusted significance 95 level of 0.05/3=0.017). However, this was not true for alpha range (Figure 1c, KW test, 96 χ 2 (133) =0.11, p=0.741). Figure 1d shows the median scalp maps (EEGLAB,26,see SI 97 Methods) of change in band-limited power across 112 bipolar electrode pairs (shown as discs) 98 for alpha, slow and fast gamma bands. We observed that stimulus-induced change in power 99 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 across all 3 bands was most prominent in the 10 electrode pairs as above. However, this change 100 in power was less in the case group compared to the control group in slow and fast gamma 101 bands (but not alpha band) as noted in Figure 1c. These trends remained similar if we compared 102 the MCI and AD groups separately with their respective controls (Supplementary Figure 1). 103 Further, these trends did not differ qualitatively if we removed participant A1 (who did not 104 have a corresponding control) from analysis. 105 To confirm these group-level results at individual level, we compared the change in 106 band-limited power for each case with the median change in power of their corresponding age 107 and gender matched controls. Figure 2 shows scatter plots of band-limited power in alpha, slow 108 and fast gamma for 19 cases (excluding A1), and the median change in power for their 109 corresponding controls. Only 3 MCI patients (and 0 AD patients) had higher slow gamma than 110 their controls. Similarly, only 2 MCI and 1 AD patients had higher fast gamma than their 111 controls. However, for alpha, the scatter was symmetrical across the identity line. This shows 112 that most of the controls had higher slow/fast gamma (and not alpha) than the cases, thus 113 corroborating the results observed in Figure 1. 114 To statistically test this, we calculated the difference between change in power for each 115 case and the median change in power of their corresponding controls, separately for alpha, slow 116 gamma and fast gamma bands. This yielded a set of 19 differences for each of the three bands.

117
The median difference was significantly less than zero for slow and fast gamma (left-tailed 118 paired Wilcoxon signed-rank test, Z=-2.35/-2.27, p=0.009/0.01 for slow/fast gamma 119 respectively), but not alpha (Z=-0.95, p=0.17). Results remained similar when we randomly 120 chose only 1 control per case (median p-values over 10,000 iterations: 0.162/0.031/0.048 for reanalysed the data after removing 4 MCI and 1 AD patients who had negative change in either 125 slow or fast gamma power (Supplementary Figure 2a). Further, to minimise error during 126 clinical diagnosis, we called for a review of the diagnosis of MCI/AD patients by an expert 127 panel that operationalised NIA-AA criteria to study setting (see SI Methods). This panel 128 revised the diagnosis of 2 MCI patients (M4 and M8) as healthy. Trends did not differ when 129 we removed these 2 patients from analysis (Supplementary Figure 2b). These results indicated 130 that both slow and fast gamma were weaker in the cases compared to the controls. 131 We found that the cases and their respective controls had comparable eye-movement 132 and microsaccade profiles, and similar pupillary reactivity to stimulus presentation (measured  Figure 4). We thus ruled out the biases introduced by these non-138 neural variables in our analyses. 139 We next tested whether power of steady-state visually evoked potentials (SSVEPs) in 140 gamma range also decreased in the case group as compared to the control group. We tested for 141 SSVEPs at 32 Hz by presenting full-screen gratings that phased-reversed at 16 Hz. 17 of the 142 20 cases participated in this study, out of which data of only 12 could be analysed (9 MCI, 3 143 AD; data from 5 cases were discarded due to noise, see SI Methods for details). These 12 cases  Figures 1c and 1d) for control and case groups, respectively. We observed a modest 151 reduction in SSVEP power at 32 Hz in the case group as compared to the control group (Figures 152 3a and 3c), but this decreasing trend was not significant (K-W Test, significance at each 153 frequency of the change in power spectra is shown in Figure 3a. Significance at 32 Hz: 154 χ 2 (87)=0.50, p=0.48). This contrasted with trends for gamma power: we reanalysed data for 155 slow and fast gamma power (as in Figure 1)  also showed a decreasing trend like in Figure 1, albeit the difference was not significant 160 (χ 2 (87)=0.85, p=0.35). Alpha followed a similar insignificant trend as in Figure 1 (χ 2 (87)=0.55, 161 p=0.45). 162 We also repeated the analysis at individual level as in Figure 2. Figure 3e shows scatter 163 plot for change in SSVEP power at 32 Hz for each of the 11 cases (participant A1 discarded) 164 and the median change in SSVEP power for the corresponding controls, same format as in 165 Figure 3. We observed that there was a lot of scatter around the identity line, suggesting that 166 SSVEP power was not less in cases compared to their controls, in contrast to slow and fast 167 gamma power. To quantify the significance of our observation, we calculated differences of 168 change in SSVEP power at 32 Hz for 11 cases and median change in SSVEP power of 169 corresponding controls. The median difference was not significantly less than 0 dB (left-tailed 170 paired Wilcoxon signed-rank test, Z=-0.22, p=0.4). Moreover, the median difference ±SD 171 (after bootstrapping over 10,000 iterations) included 0 dB mark (inset in Figure 3e), suggesting 172 that these differences were not significantly different from 0 dB. The trends for alpha/slow/fast 173 gamma for these 11 cases and their respective controls were comparable to that discussed in 174 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Figure 2). To conclude, change in SSVEP 175 power at 32 Hz for cases was comparable to that of their controls, like alpha but unlike slow 176 and fast gamma oscillations. 177 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Discussion 178 Stimulus-induced change in power of both narrow-band slow and fast gamma 179 oscillations reduced in elderly patients with clinical MCI/AD compared to their age and gender 180 matched healthy controls. We removed or ruled out possible biases due to peripheral ocular 181 factors or overall baseline noise of the PSDs to further strengthen the results. In contrast to 182 gamma, we did not find any significant reduction in stimulus-induced alpha suppression or 183 SSVEP at 32 Hz in cases.  Our sample is a representative of urban population in India as we adopted community-192 based sampling instead of hospital-based sampling. Importantly, out of the 257 participants 193 that we collected data from (247 used for analysis plus 10 participants whose data was noisy 194 and thus rejected, see SI Methods), there were 15 MCIs (5.8%) and 6 AD patients (2.3%).

195
These figures match closely to the previously reported prevalence of MCI and AD in India (22,196 35,36). The main strength of our study is that within this sample, most cases (70%) had MCI,197 a condition that is conceptualized as intermediate stage between normal aging and AD. Criteria 198 used to diagnose MCI are not strong and hence there is a need for a valid biomarker. Our study 199 highlights the potential use of gamma oscillations in EEG in that direction. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020 Moreover, we had limited our analyses to sensor (electrode) level instead of 202 reconstructing the neural sources and performing analyses at that level. This has allowed us to 203 present a diagnostic technique that is easy to replicate in a clinical setting. Furthermore, as our 204 metrics were derived directly from neural activity, these could serve as a more objective

215
Age is as a common factor that influences both macroscopic structure (45-47) as well as 216 gamma power and frequency (17).

217
Our main aim in this study was to examine the potential of gamma as a biomarker.

218
However, as structural changes in AD brain are more evident and drastic compared to  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020 Some investigators have suggested neuroprotective effects of entraining neural 226 oscillations using flickering light/sound in gamma frequency range (analogous to our SSVEP 227 paradigm), in rodent models of AD (14,15,48). However, we did not find any significant trend 228 for SSVEP at 32 Hz in cases compared to controls, unlike our observations with narrow-band 229 gamma. There could be several reasons for these differences. First, it is possible that 230 entrainment of neural oscillations to visual stimulation in gamma frequency range gets 231 deranged only in advanced stages of AD (as in the rodent models of 13). It may thus have 232 therapeutic benefit but may not reflect as abnormal on testing early on (as in our case). Second, 233 as discussed in the SI Methods, SSVEP study was always done at the end of the experiment in 234 our study and the total number of stimulus repeats were much less than the gamma study.

235
To conclude, stimulus-induced change in visual narrow-band gamma power has the 236 potential to be a simple, low-cost, easy to replicate and objective biomarker for diagnosis of (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020.  Table 1). 5 out of the 6 AD patients were directly referred to the study by the neurologist.

250
Diagnosis of all MCI/AD patients was further reviewed by a panel of 4 experts for consensus 251 (see SI Methods). We discarded data of 10 participants due to noise (see Artifact Rejection 252 section in SI Methods). We were thus left with 227 healthy participants, 14 MCI and 6 AD 253 patients for analysis. For the purpose of this study, we called the MCI/AD patients as cases 254 (N=20) and their respective age and gender matched healthy participants as controls. were presented for 800 ms (with interstimulus interval of 700 ms) at one of three spatial 267 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint frequencies (SFs): 1, 2, and 4 cycles per degree (cpd) and four orientations: 0°, 45°, 90° and 268 135° for the Gamma experiment, chosen pseudorandomly. We also presented gratings (with 269 one SF and orientation combination that showed high change in gamma power for each 270 participant during preliminary analysis performed for Gamma experiment) counter-phasing at 271 16 cycles per second (cps), to test for SSVEPs at 32 Hz in the SSVEP experiment. Both the 272 experiments were performed in a single recording session lasting ~30 minutes. We recorded 273 64-channel EEG using active electrodes (international 10-10 system of electrode placement) 274 and BrainAmp DC (Brain Products GmbH). We also recorded eye data using an infrared eye-   (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.    (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.   (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. and fast gamma (right). Color of disks represents change in power in respective frequency 495 bands. Electrode groups used for calculation of band-limited power are enclosed in black.

497
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Number of subjects (no. of females in parenthesis)
Total recruited 236 (104)  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

540
Consensus diagnosis 541 As in Table 1, our major criteria to diagnose MCI/AD were clinical history and 542 CDR which was administered by a single clinician (neurologist or psychiatrist). However, to 543 minimize observer bias, we followed the recommendations made in SAGES study (52)   When they could not achieve at a consensus for any subject in the first instance (6 subjects), 554 they used Delphi method (54, 55) till all of them agreed upon the diagnosis. Briefly, the 555 members were informed of the discrepancy within the panel, who then discussed among 556 themselves and rated again. This process was iterated till all 4 members came to a consensus.

557
Although they used this stringent approach to confirm the previous diagnoses, they reclassified 558 as healthy only 2 subjects previously classified as MCI (M4 and M8, see Supplementary Figure   559 1). They confirmed and retained initial diagnosis for rest of the 12/14 MCI and 6/6 AD patients.

560
The analyses results for these patients are presented in Supplementary Figure 2b. 561 562 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint Experimental setup and task 563 Experimental setup, EEG recordings and analysis were same as what we had described 564 in our previous study (17). Briefly, we recorded raw EEG signals from 64 active electrodes 565 using BrainAmp DC (Brain Products GmbH) according to the international 10-10 system, 566 referenced online at FCz. We filtered raw signals online between 0.016 Hz and 1000 Hz and 567 sampled at 2500 Hz. We rejected electrodes whose impedance was more than 25 KΩ (4.0% 568 and 2.5% for healthy subjects and cases respectively). Impedance of the final set of electrodes 569 was 5.5±4.2 and 5.2±4.3 KΩ for healthy subjects and cases respectively.

570
All subjects sat in a dark room in front of an LCD screen with their head supported by a 571 chin rest. The screen (BenQ XL2411, resolution: 1280 x 720 pixels, refresh rate: 100 Hz) was 572 gamma-corrected and placed at a mean±SD distance of 58.1±0.8 cm from the subjects (53.8-573 61.0 cm) according to their convenience (thus subtending a width of at least 52º and height of 574 at least 30º of visual field for full-screen gratings). We calibrated the stimuli to the viewing 575 distance in all cases.

576
Subjects performed a visual fixation task. Every trial started with the onset of a fixation 577 spot (0.1°) at the center of the screen on which the subjects had to maintain fixation. After an 578 initial blank period of 1000 ms, a series of stimuli (2 to 3 sinusoidal luminance gratings 579 presented full screen at full contrast) were randomly shown for 800 ms each with an inter-580 stimulus interval of 700 ms. For the main "Gamma" experiment, these were presented at three 581 spatial frequencies (SFs): 1, 2, and 4 cycles per degree (cpd) and four orientations: 0°, 45°, 90° 582 and 135°. Subjects performed this task in 2-3 blocks (total 543 blocks across 257 subjects) 583 during a single session according to their comfort. 584 We also tested 32-Hz SSVEPs on these subjects in the SSVEP experiment. Gratings (with 585 one SF and orientation combination that showed high change in slow and fast gamma power 586 for each subject during preliminary analysis performed during the session) counter-phased at 587 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint 16 cycles per second (cps) in a similar stimulus presentation paradigm as described above, 588 randomly interleaved with static gratings of the same SF and orientation. A few subjects 589 enrolled during the initial phase of the study (2 MCI, 1 AD and 1 healthy subject) did not 590 undergo this experiment. Further, we considered only those subjects for analysis in SSVEP 591 experiment who had analyzable data for the Gamma experiment (see Artifact Rejection section 592 below). This gave us a total of 221/12/5 subjects (99/2/2 females) for healthy/MCI/AD 593 categories for the SSVEP experiment. 594 We presented each stimulus ~30-40 times for both the Gamma and SSVEP experiments 595 according to the subjects' comfort and willingness. The entire recording session lasted for ~30 596 minutes. Gamma experiment lasted for ~25 minutes, with 1-2 short breaks (for 3-5 minutes) 597 between blocks. This was followed by the SSVEP experiment (for ~5 minutes) completed in 598 one block. Unless otherwise stated, stimulus presentation of a particular orientation and spatial 599 frequency is referred to as a "stimulus repeat" in this paper.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Artifact rejection 613 We used a pipeline to reject artifact-containing data as described in Murty et al. (17). 614 Briefly, we applied a repeat-wise thresholding process on both time-domain waveforms and 615 multi-tapered PSD (between -500 ms to 750 ms of stimulus onset) to select bad repeats across 616 electrodes. We discarded those electrodes that had more than 30% of all repeats marked as bad, 617 and subsequently labelled any repeat as bad if it occurred in more than 10% of total number of 618 remaining electrodes. We next discarded those electrodes that had PSD slopes (calculated in 619 56 Hz to 84 Hz range as described in 17) less than 0. Finally, we discarded any block that did were analyzed. We then pooled data across all good blocks for each subject for final analysis.

626
Those subjects who did not have any analyzable blocks (9/236 and 1/21 for healthy subjects 627 and cases respectively) were discarded from further analysis. The total number of repeats that 628 were finally analyzed were 276.2±87.2 for healthy subjects and 246.9±72.8 for cases.

629
A similar procedure was used for SSVEP experiment yielding 30.2±6.9 and 30.7±8.4 630 repeats for healthy subjects and cases respectively. Note that this experiment was always done 631 towards the end, and therefore the signal quality could be poorer than the Gamma experiment.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 EEG data analysis 638 For all analyses we re-referenced data at each electrode offline to its neighboring 639 electrodes (bipolar reference). We thus obtained 112 bipolar pairs out of 64 unipolar electrodes 640 (17). We considered the following bipolar electrodes for analysis: PO3-P1, PO3-P3, POz-PO3,  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020 at lower frequencies within the band and diverge at higher frequencies, which was the case in 707 the slow gamma range for both MCI and AD groups compared to their healthy controls, the 708 total change in power in the band may not be significantly different. Therefore, our results 709 could be improved by customizing the low frequency limit of the gamma band for each subject, 710 as well as choosing only electrodes that show stronger gamma. For example, taking slow 711 gamma range as 24-30 Hz improved p-values for both MCI (K-W test, χ 2 (89)=3.50, p=0.06) 712 and AD (χ 2 (55)=6.74, p=0.009). We have refrained from such customization here because we 713 wanted to study the efficacy of a simple and subject-independent computational procedure, but 714 such data-driven subject specific optimization holds promise for improving the efficacy of a 715 gamma-based biomarker. Moreover, there was a wide range of age of the MCI/AD cases for 716 group-level analyses. As gamma power was shown to depend on age (17), this could increase 717 the variability in our data and hence adversely affect the observed significance. This problem (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. and 5 AD; and their respective controls) whose diagnosis was confirmed by all 4 experts after 733 the consensus diagnosis exercise (see SI Methods). Patients M4 and M8 were removed from 734 analysis thus. Left-tailed paired Wilcoxon signed-rank test: alpha, Z=-1.09, p=0.14; slow 735 All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint gamma, Z=-2.08, p=0.019; and fast gamma Z=-1.99, p=0.023. Participant A1 (who also had 736 change in both slow and fast gamma power >0 dB and whose diagnosis was confirmed during 737 consensus diagnosis exercise) was excluded from both analyses above as he had no healthy 738 control.

740
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint across all subjects for each group is also indicated at the bottom of the panels. MCI/AD cases 754 had similar microsaccade rates (also seen in panel 3a) and main sequence plots compared to 755 their healthy controls. c) Scatter plots showing change in power for cases (abscissa) and median 756 change in power for corresponding healthy controls (ordinate) in alpha (right), slow gamma 757 (middle) and fast gamma (left) bands. Same format as in Figure 2, but stimulus repeats 758 containing microsaccades have been discarded from analysis. Trends discussed in Figure 2  medians and error bars indicate ±SD of median of 10,000 bootstrapped samples. We did not 764 find any significant difference between the MCI/AD and control groups in pupil reactivity as 765 seen in the plot. These results ruled out potential biases due to ocular factors that could have 766 influenced the results discussed in the text (Figures 1 and 2). (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020 PSDs for MCI/AD cases were overlapping with those of the control subjects. b) Scatter plots 775 showing baseline absolute power (calculated in -500 -0 ms of stimulus onset) for each of the 776 19 cases (abscissa) and median change in power for corresponding healthy controls (ordinate) 777 in alpha (right), slow gamma (middle) and fast gamma (left) bands. c) Scatter plots showing 778 baseline PSD slopes for each of the 19 cases (abscissa) and median change in power for 779 corresponding healthy controls (ordinate). Frequency range considered for each scatter plot is (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.24.20139113 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.