Individualized Prediction of Task Performance Decline Using Pre-Task Resting-State Functional Connectivity

As a common complaint in contemporary so-ciety, mental fatigue is a key element in the deterioration ofthe daily activities known as time-on-task (TOT) effect, mak-ing the prediction of fatigue-related performance declineexceedingly important. However, conventional group-levelbrain-behavioral correlation analysis has the limitation of generalizability to unseen individuals and fatigue predic-tion at individual-level is challenging due to the significantdifferences between individuals both in task performanceefficiency and brain activities. Here, we introduced a cross-validated data-driven analysis framework to explore, for thefirst time, the feasibility of utilizing pre-task idiosyncraticresting-state functional connectivity (FC) on the predictionof fatigue-related task performance degradation at indi-vidual level. Specifically, two behavioral metrics, namely Delta RT (between the most vigilant and fatigued states) andTOTslopeover the course of the 15-min sustained attentiontask, were estimated among three sessions from 37 healthysubjects to represent fatigue-related individual behavioralimpairment. Then, a connectome-based prediction modelwas employed on pre-task resting-state FC features, iden-tifying the network-related differences that contributed tothe prediction of performance deterioration. As expected,prominent populational TOT-related performance declineswere revealed across three sessions accompanied withsubstantial inter-individual differences. More importantly,we achieved significantly high accuracies for individual-ized prediction of both TOT-related behavioral impairmentmetrics using pre-task neuroimaging features. Despite thedistinct patterns between both behavioral metrics, the iden-tified top FC features contributing to the individualized pre-dictions were mainly resided within/between frontal, tem-poral and parietal areas. Overall, our results of individual-ized prediction framework extended conventional correla-tion/classification analysis and may represent a promisingavenue for the development of applicable techniques thatallow precaution of the TOT-related performance declinesin real-world scenarios