نوع مقاله : علمی-پژوهشی
نویسندگان
اذربایجان شرقی،شهرجدیدسهند،دانشگاه صنعتی سهند،دانشکده مهندسی پزشکی
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Functional magnetic resonance imaging (fMRI) is a crucial tool for investigating brain activity. However, analyzing fMRI data presents significant challenges due to the complex temporal nature of the signals and uncertainties in the algorithms used. Classical Independent Component Analysis (ICA) algorithms, such as FastICA, often struggle with high false positive rates and unstable results because they rely on the strict assumption of complete statistical independence. This study aims to comprehensively compare three ICA-based algorithms: FastICA, Entropy Bound Minimization (ERBM), and Semi-Blind Spatial ICA (SBS-ICA). The objective is to assess how different statistical assumptions and prior information affect the quality of component separation in fMRI data and the accurate identification of brain activation regions. Evaluations were conducted using component numbers of 40, 50, 60, and 70. The results revealed that the SBS-ICA algorithm, which benefits from spatial prior information, demonstrated the best performance with an area under the ROC curve (AUC) of 0/999, a high correlation of 0/89, and the lowest number of spatial false positives. The ERBM algorithm, which models temporal correlations, outperformed FastICA, showing a lower mean squared error (MSE = 0/079) and more stable correlation values. In contrast, FastICA exhibited the weakest performance among the three algorithms. These findings highlight the advantages of ICA-based guided methods and emphasize the significance of incorporating task modeling for accurate analysis of fMRI data.
کلیدواژهها [English]
)