SuperAgers are determined by the scores of global cognitive function. This ensured that the participants are divided by a measure relevant to the current study without directly biasing the results toward the domain of interest.

The cases with the scores of cognitive performance within 1.5 standard deviation (SD) of age and education matched normative values derived from our cohort study, of which presented with HK MoCA score greater than 26 and CDR score equal to 0

NCD patients are defined as: evidence of modest decline in one or more cognitive domains, which is set as ≥ 1.5 SD below the age- and education-adjusted normative scores; no interference with independence in everyday activities; and no comorbid major psychiatric disorders.

Using comprehensive cognitive tests to quality the domain-specific cognitive changes during ageing.

The Montreal Cognitive Assessment Hong Kong version (HK MoCA) is a validated and widely used screening assessment for detecting cognitive impairment, which measures an array of cognitive functions such as visuospatial, memory, language and attention.HK MoCA score less than 26 indicates cognitive impairments.

Attention Network Test (ANT). The ANT paradigm (Download from: https://www.sacklerinstitute.org/cornell/assays\_and\_tools/ant/jin.fan/), as a computerized test, was used and performed by E-Prime 3.0 (Psychology Software Tools, PA, USA). Within ANT paradigm, there are four types of cue: no cue, center cue, double cue, and spatial cue; and three types of flanker: neutral, congruent, and incongruent. In a given trial, a central cross-fixation point presents for 400 to 1,600 ms (randomized), subsequently is replaced for 100 ms by one of four warning cues. The target, a central arrow could appear above or below the cross-fixation and is surrounded by two flankers on each side.

Word list learning test (WLLT) consisting of fifteen semantically non-associated words that is presented consecutively over three free trials of immediate recall (i.e., learning ability), a 20-min delayed recall (to prevent recency effects) and followed by yes/no recognition (i.e., recognition memory). Based on the proxies we developed in our previous work, components of learning and memory function will be evaluated and compared across groups, including working memory capacity (WMC), forgetting rate, delayed recall, initial learning, best learning, learning over trials (LOT) and learning ratio.