سامانه همایش‌ها TSTA | ثبت‌نام و ارسال مقاله کنفرانس‌ها و کنگره‌های ملی

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Dynamic Classification of Resting-State EEG Using Adaptive Functional Connectivity in Mild Traumatic Brain Injury

نویسندگان :

Farzaneh Manzari¹ Peyvand Ghaderyan²

1- دانشگاه صنعتی تبریز(سهند) 2- دانشگاه صنعتی تبریز(سهند)

کلمات کلیدی :

Adaptive Intrinsic Warped Connectivity،Dynamic connectivity feature،Functional Connectivity،Mild Traumatic Brain Injury،Recurrent Neural Network

چکیده :

Mild Traumatic Brain Injury (mTBI) is a common and potentially life-threatening medical condition that can lead to long-lasting physical and cognitive impairments and accounts for 90% of all TBIs. Brain injury is associated with disrupted connectivity resulting from damage to the axons. This study investigated brain connectivity using resting-state electroencephalogram (EEG) signals in patients with mTBI. Hence, a novel hybrid approach, namely Adaptive Intrinsic Warped Connectivity (AIWC), has been proposed to examine functional connectivity, based on empirical mode decomposition, analytical representation, and dynamic warping techniques. AIWC effectively reveals dynamic, nonlinear coupling patterns that reflect the brain’s spontaneous organization and transition states. An analysis has been conducted to compare the detection performance of the proposed connectivity features across symmetrical or asymmetrical channel pairs. The study also examined the effects of several sessions after injury, as well as the utilization of amplitude and phase information through the Recurrent Neural Network as a dynamic mapping procedure, and has been evaluated using a 10-fold cross-validation technique. The suggested algorithm has been tested using resting-state EEG signals from the initial session, as well as at two- and four-month follow-ups with 52 individuals diagnosed with mTBI and 31 healthy controls. The results obtained show a high average accuracy rate of 99.22%. The analysis has also demonstrated the proposed feature's higher performance across the frontal lobe, frontal-parietal, frontal-temporal, and frontal-occipital regions.