<p>Numerous neuroimaging studies have explored brain structural biomarkers to distinguish mild cognitive impairment (MCI) from cognitively normal (CN) individuals using cross-sectional data. We further examined alterations in brain cortical thickness within the visual cortices over time, alongside neuropsychological tests, to predict the progression to MCI in a cohort of older adults. Our dataset included 216 CN controls (non-converters) and 32 CNc subjects (converters) at baseline. After the follow-up period, 216 CN subjects remained stable (CNs), while 32 CNc subjects progressed to MCI. We utilized T1-weighted brain images for neuroimaging analysis, employing a three-dimensional convolutional neural networks-based deep learning algorithm to assess cortical thickness in the visual cortices at baseline and follow-up. Demographic characteristics and cortical thickness were compared using a two-sample t-test, and subsequent classification models were developed. At baseline, no significant differences in neuropsychological scores were observed between the CN and CNc groups. However, during follow-up, a significant difference emerged between the CNs and MCI groups. Notably, the CNc group, compared with the CN group at baseline, and the MCI group, compared with the CNs group at follow-up, exhibited significant cortical thinning in the visual cortices at their respective time points. Additionally, the classification models, incorporating cortical thickness and neuropsychological scores, demonstrated acceptable and robust performance. Specifically, the area under the curve values were 0.62 − 0.67 for CN vs. CNc at baseline and 0.93 for CNs vs. MCI at follow-up. Our findings indicate that cortical thinning in the visual cortices and the performance of classification models hold significant potential for identifying an increased risk of MCI among older adults.</p>