Event-related potential markers of subjective cognitive decline and mild cognitive impairment during a sustained visuo-attentive task

Abstract

Subjective cognitive decline (SCD), mild cognitive impairment (MCI), and Alzheimer's disease stages lack well-defined electrophysiological correlates, creating a critical gap in the identification of robust biomarkers for early diagnosis and intervention. In this study, we analysed event-related potentials (ERPs) recorded during a sustained visual attention task in a cohort of 178 individuals (119 SCD, 40 MCI, and 19 healthy subjects, HS) to investigate sensory and cognitive processing alterations associated with these conditions. SCD patients exhibited significant attenuation in both sensory (P1, N1, P2) and cognitive (P300, P600, P900) components compared to HS, with cognitive components showing performance-related gains. In contrast, MCI patients did not show a further decrease in any ERP component compared to SCD. Instead, they exhibited compensatory enhancements, reversing the downward trend observed in SCD. This compensation resulted in a non-monotonic pattern of ERP alterations across clinical conditions, suggesting that MCI patients engage neural mechanisms to counterbalance sensory and cognitive deficits. These findings support the use of electrophysiological markers in support of medical decision-making, enhancing personalized prognosis and guiding targeted interventions in cognitive decline.

Keywords: EEG; Event-related potentials; Mild Cognitive Impairment (MCI); Subjective Cognitive Decline (SCD).

Fig. 1

Visual abstract. (A) Monotonic decreasing relationship between age and cognitive decline, modulated in its course by the presence of SCD or MCI pathology (or eventually AD). Dashed line traces the pathological deviation from healthy path line. (B) Performance on the visuo-attentive task 3-CVT following a monotonic decreasing course according to the severity of the pathology. (C) Experimental EEG signal recording setting while participants were performing the 3-CVT task. (D) ERP dynamics extracted from occipital and central channels separated by clinical condition, where occipital channels probed the encoding phase of the stimulus, and the central channels probed the decision-making phase of the stimulus. (E) Non-monotonic (V-shape) trend of ERP features reflecting increased recruitment of neural resources by MCI patients. Dashed line traces the hypothetical monotonic trend which has been altered by the features values.

Fig. 2

Behavioural outcomes across conditions. (A) F-measure [a.u.]. (B) Reaction time [s]. (C) Accuracy [%]. To the right of (C) is the Cumulative Density Function of Accuracy dichotomised into 'low' and 'high' levels with reference to the median of the aggregated conditions. Pairwise statistics based on Kruskal-Wallis H test with p-value corrected by Bonferroni's method (alpha = 0.05). P-value annotation legend: ns: 0.05 < p <= 1, *: 0.01 < p <= 0.05, **: 0.001 < p <= 0.01, ***: 0.0001 < p <= 0.001. Colour code: HS (blue), SCD (orange), MCI (red), Low-performance subjects (green) and High-performance subjects (cyan). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

ERP dynamics across conditions. (A). ERP computed as average of signals in the cluster of occipital channels (PO7, PO8, O1, Oz, O2) representative of the encoding phase of the stimulus. (B) ERP computed as average of signals in the cluster of central channels (FC1, FCz, FC2, C1, Cz, C2) representative of the decision-making phase regarding the stimulus. Both panels (A) and (B): Bold representation is the overall mean within each group and shading is the standard deviation. The measures on top are the instantaneous H-statistic of the Kruskal-Wallis test and the associated p-value corrected by Bonferroni's method (alpha < 0.05). Temporal instants associated with a p < 0.05 are highlighted with a vertical violet bar. P1/N1/P2 and P300/600/900 labels stand for the name of the event-related potentials relative to the encoding phase and decision-making phase, respectively. Colour code: HS (blue), SCD (orange), MCI (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Aggregated feature types normalized by HS values. The feature types (latency, peak, integral) aggregated the potentials (P1/N1/P2/P300/P600/P900) and are normalized by HS mean (0 % variations means the values are closed to HS average value; if % is > 0/<0 means positive/negative percentage deviation from HS mean). Each panel showed bar and line plots to highlight mono/non-monotonic trend. (A) ERP latency. (B) ERP peak. (C) ERP integral. Pairwise statistics based on Kruskal H test with p-value corrected by Bonferroni's method (alpha = 0.05). P-value annotation legend: ns: 0.05 < p <= 1, *: 0.01 < p <= 0.05, **: 0.001 < p <= 0.01, ***: 0.0001 < p <= 0.001. Colour code: SCD (orange), MCI (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

Radar plots of normalized integral features. The graphs show the percentage deviations of each ERP integral feature from the HS case of the SCD (A) and MCI (B) conditions, and the percentage deviation of SCD from MCI (C). The hexagonal box on each radar plot represents the percentage limit of 0 % (HS for A-B and MCI for C), the positive or negative deviations of which show the variations from that reference. The numbers on the corner points of the polygon indicate the deviation in percentage terms for each integral characteristic. The legend shows the average of the percentage changes of the integrals (indicated by ). The pairwise statistic based on the Kruskal H test with p-value corrected by Bonferroni's method (alpha = 0.05) is close to each percentage pair. Annotation legend P-value: ns: 0.01 < p <= 1, ∼: 0.05 < p < 0.1, *: 0.01 < p <= 0.05, **: 0.001 < p <= 0.01, ***: 0.0001 < p <= 0.001. Colour code: HS (blue), SCD (orange), MCI (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6

ERP dynamics across performance in SCD. (A). ERP computed in the cluster of occipital channels (PO7, PO8, O1, Oz, O2) representative of the encoding phase of the stimulus. (B) ERP computed in cluster of central channels (FC1, FCz, FC2, C1, Cz, C2) representative of the decision-making phase regarding the stimulus. Both panels (A) and (B): Bold representation is the overall mean within each group and shading is the standard deviation. The measures on top are the instantaneous H-statistic of the Kruskal-Wallis test and the associated p-value corrected by Bonferroni's method (alpha < 0.05). Temporal instants associated with a p < 0.05 are highlighted with a vertical violet bar. P1/N1/P2 and P300/600/900 labels stand for the name of the event-related potentials relative to the encoding phase and decision-making phase respectively. Colour code: low performance subjects (green), high performance subjects (cyan). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7

ERP dynamics across performance in MCI. (A). ERP computed in the cluster of occipital channels (PO7, PO8, O1, Oz, O2) representative of the encoding phase of the stimulus. (B) ERP computed in cluster of central channels (FC1, FCz, FC2, C1, Cz, C2) representative of the decision-making phase regarding the stimulus. Both panels (A) and (B): Bold representation is the overall mean within each group and shading is the standard deviation. The measures on top are the instantaneous H-statistic of the Kruskal-Wallis test and the associated p-value corrected by Bonferroni's method (alpha < 0.05). Temporal instants associated with a p < 0.05 are highlighted with a vertical violet bar. P1/N1/P2 and P300/600/900 labels stand for the name of the event-related potentials relative to the encoding phase and decision-making phase respectively. Colour code: low performance subjects (green), high performance subjects (cyan). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 8

Encoding FC computed across clinical conditions and ATN classification. (A) Boxplot of clinical conditions. (B) Boxplot of clinical conditions stratified by ATN status. Topographic maps left to (A) are indicative of the scalp potentials spread during the encoding process (0–200 ms) as the FC is over/under the FC median. Colormaps are in voltage range −1/+1 [µV]. Higher FC means higher similarity between the occipital seed and other channels, indicative of increased extra-occipital scalp recruitment. Pairwise comparisons between groups are shown with p-values from the Kruskal-Wallis test. Significant p-values are marked with asterisks: * (p < 0.05), ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001).