老年主動推理認知訓練與相關神經機制

Abstract

已知老化會造成大腦認知功能的衰退，因此為維持大腦認知功能而產生了各式的認知訓練。一般來說，傳統訓練的方式在於反覆引導參與者使用特定的認知功能，並達到於相對應的認知測驗分數上得到更高分的表現。然而此訓練的效果是否能轉移至其他未受到直接訓練所刺激的認知能力有待驗證。本研究著重研究大腦與外界環境間的交互關係，當大腦接收外界刺激，會更新原有信念並影響下次對外界的推理。推理可以區分為主動與被動，我們推論使用主動推理能力的同時需要整合多種認知能力，因此我們設計一認知訓練，參與者需要使用主動推理來完成任務，不同於過去的認知訓練針對單一認知能力重複刺激，本訓練著重引導參與者持續主動推理完成任務。本訓練使用樂高公司拼裝之機器車與電腦軟體，包含一系列目標導向型任務讓參與者完成，並依完成方式區分為樂高實驗組與控制組。此外，欲驗證訓練效果，在訓練前後使用一顏色規則推理的測驗請參與者於功能性磁振造影過程中完成，該測驗同時包括主動推理與被動推理的單元，參與者必須依靠主動或被動推理找出隱藏在各個組合背後的分類規則。對於樂高實驗組而言，在主動與被動推理的行為結果差異不大。神經活動方面，於主動推理部分之前後測結果顯示由訓練前與視覺推理相關的後葉，至訓練後更多關於推理有關的額葉訊號產生。另外樂高控制組於被動推理部分顯示視覺運動區的活化，且於後測結果顯示差異不大的結果。樂高實驗組前後測之神經心理測驗結果上得到些微的進步，然而此結果尚無法直接驗證此訓練效果是否轉移。綜合來說，本研究旨在設計一短時間且小規模的認知訓練，針對訓練提升主動推理能力，由初步的實驗結果觀察到腦區神經活動之變化加以驗證主動推理確實能改變老年人大腦與環境之交互作用。

Traditional cognitive training for older adults focuses on continuous practice of specific abilities like memory, processing speed and reasoning. While such approaches usually result in improvements in the trained cognitive abilities, transfer to other non-trained abilities is limited. In the study, we evaluate the efficacy of a cognitive training approach in older adults based on how the brain interacts with the external environment to form inferences. Inferencing refers to how the brain updates its beliefs and uses them to predict the future. Inferencing can be active or passive, which involves doing actions to verify predictions or just observing events. Critically, inferencing requires the brain to integrate multiple cognitive functions so that training in active inference might transfer. We designed a 6-hr Lego Robot Programming (LRP) cognitive training intervention for participants to engage inferential processing. 24 older adult participants (mean age = 66.6 yrs, SD = 5.48 yrs) were separated into the experimental active inference group (N = 12) and the control passive inference group (N = 12), which dissociated according to whether participants could control the robot directly or not. To evaluate the training effect, we applied a Visual Rule Inference Task (VRIT) in functional magnetic resonance imaging (fMRI) session during pre- and post-training assessments. VRIT behavioral performances were similar for both groups. However, VRIT neural responses shifted from visuoparietal to the middle cingulate area from pre- to post-training in the active inference group but remained in visuo-parietal areas for the passive inference group. Our findings reveal functional brain differences during active and passive inference about the environment. Moreover, we demonstrate somewhat meaningful changes in older adult neural engagement after a short 6-hr cognitive training protocol for active inference. Future analysis will consider whether LRP training transfers to other cognitive abilities.