A Frequency-constrained Spectrum Difference Mapping Framework for Decoding Brain Activity from Functional Magnetic Resonance Imaging Data

Abstract

Many studies have shown that spontaneous low-frequency oscillation is an intrinsic attribute of human brain activity based on the resting-state blood oxygen level-dependent (BOLD) functional Magnetic Resonance Imaging (fMRI) technology. Amplitude of low-frequency fluctuations (ALFF) is an effective way to capture the low-frequency fluctuation and has a hugely wide range of applications in mental disorders, neurological diseases, occupational neuroplasticity, etc. Such approaches, however, needs further improvement in two problems: low sensitivity to low-frequency signals; noise signal interference. Based on this, this paper proposes a frequency-constrained spectrum difference mapping framework (SDMF). A frequency domain is transformed through fast Fourier transform (FFT) and divided by designating Frequency(low), Frequency(mid), and Frequency(high). Then, spectrum difference value (SDV) is calculated between the two regions as the characterization value of brain activity state. Through our experimental results, we propose that SDMF can achieve the noise reduction effect, and it is the same as the region of the spontaneous active state in the traditional ALFF. In our method, it also showed that SDMF with different metrics has achieved the suppression of the Temporal, Lingual, and enhancement of the Occipital region. All in all, SDMF is a basic framework to analyze the band constraints, and the traditional ALFF can also be included as a special mode.