نوع مقاله : علمی-پژوهشی
نویسندگان
اذربایجان شرقی، شهرجدیدسهند، دانشگاه صنعتی سهند، دانشکده مهندسی پزشکی
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Clustering analysis of functional Magnetic Resonance Imaging (fMRI) time series is a widely used method for non-invasively identifying and mapping active brain regions in response to specific tasks. However, fMRI clustering faces significant challenges due to the imbalance between active and inactive voxels, often leading to a high number of false positives and the creation of noisy, scattered activation maps. This study introduces a novel two-stage method designed to control the false positive rate and enhance the accuracy and quality of detecting active brain regions. In the first stage, the method establishes a statistical threshold for signal discrimination using randomization-based inference. The second stage removes noisy and isolated voxels by applying a spatial threshold based on three-dimensional neighborhood criteria. This control strategy was tested with Neural Gas (NG) and Growing Neural Gas (GNG) clustering algorithms during an experimental analysis of auditory fMRI data. The performance of the methods was evaluated using the Dice coefficient, spatial coherence and Calinski Harabasz Index metrics. The results showed that the proposed control method significantly enhanced performance, increasing the Dice coefficient for the GNG algorithm from 0/30 to 0/65 and improving compactness by reducing the spatial coherence value from 25/62 to 14/63. The Calinski Harabasz index also increased from 2/703 to 1/1192. These results indicate that combining the GNG algorithm with this control method leads to greater stability and reliability in identifying active neural regions, providing a more effective approach for analyzing fMRI data.
کلیدواژهها [English]
)