Exploring neural activity in inflammatory bowel diseases using functional connectivity and DKI-fMRI fusion

Highlights

• fMRI demonstrated altered brain function networks in patients with inflammatory bowel disease.

• fMRI and DKI data were fused using graph Fourier transform to obtain more recognizable features.

• The feature classification results of multimodal MRI fusion are significantly better than traditional graph theory methods.

Abstract

Although MRI has made considerable progress in Inflammatory bowel disease (IBD), most studies have concentrated on data information from a single modality, and a better understanding of the interplay between brain function and structure, as well as appropriate clinical aids to diagnosis, is required. We calculated functional connectivity through fMRI time series using resting-state functional magnetic resonance imaging (rs-fMRI) and diffusion kurtosis imaging (DKI) data from 27 IBD patients and 29 healthy controls. Through the DKI data of each subject, its unique structure map is obtained, and the relevant indicators are projected onto the structure map corresponding to each subject by using the graph Fourier transform in the grasp signal processing (GSP) technology. After the features are optimized, a classical support vector machine is used to classify the features. IBD patients have altered functional connectivity in the default mode network (DMN) and subcortical network (SCN). At the same time, compared with the traditional brain network analysis, in the test of some indicators, the average classification accuracy produced by the framework method is 12.73% higher than that of the traditional analysis method. This paper found that the brain network structure of IBD patients in DMN and SCN has changed. Simultaneously, the application of GSP technology to fuse functional information and structural information is superior to the traditional framework in classification, providing a new perspective for subsequent clinical auxiliary diagnosis.

Introduction

Inflammatory bowel disease(IBD), including Crohn's disease (CD) and ulcerative colitis (UC), is a chronic inflammatory disease [1]. Studies have demonstrated that IBD is an immune-mediated disease and a comorbidity of mental health. It has been found that 16–30% IBD patients have a prevalence of depression during remission and 34–60% during active disease [2], [3]. There is substantial evidence of a bi-directional interaction between depression and IBD, with depression leading to increased hospitalization rates and reduced quality of life prior to IBD onset. On the other hand, IBD can make depression worse and aggravate the course of the disease [4]. The critical way of two-way communication involves the microbial-gut-brain axis, two-way communication, and regulatory system between the gut and the central nervous system (CNS) which affects the pathogenesis of depression and IBD [5]. The clinical process of IBD alternates between remission and recurrence, which seriously affects patients' mental health. The patient's quality of life is significantly reduced, resulting in both lost productivity and huge medical bills.

Resting-state functional magnetic resonance imaging (rs-fMRI) is a powerful tool for measuring spontaneous neural communication in the human brain [6]. Using rs-fMRI data to construct brain functional connectivity networks can reveal the pathological basis of brain diseases and develop biomarkers [7]. Meanwhile, rs-fMRI can detect the brain's random coactivation pattern, reflecting the internal functional connections between brain regions [8]. Compared with fMRI, diffusion kurtosis imaging (DKI) is a new diffusion imaging technology based on traditional diffusion tensor imaging (DTI), reflecting the non-Gaussian diffusion characteristics of water molecules in biological tissues. It can reflect the complexity of tissue microstructure and the corresponding pathological changes of diseases [9]. DKI technology based on non-Gaussian diffusion theory was first proposed by Professor Jensen in 2005 [10]. After continuous development and improvement, DKI has been widely used to study various system diseases and has shown significant clinical value [11], [12]. In recent years, functional and structural brain magnetic resonance imaging (fMRI and sMRI) technology has been widely used in the computer-aided diagnosis of IBD. More and more studies have shown that dysfunction of the central nervous system (CNS) and brain-gut axis (BGA) is one of the main factors in the pathogenesis of IBD [13], [14]. Functionally, the blood oxygen level dependent (BOLD) signals in IBD patients' amygdala, thalamus, and cerebellum were significantly reduced compared with healthy controls [15]. At the same time, IBD patients showed local and overall brain function network topology pattern disorder, including the decline of node graph theory indicators of subcortical network, sensorimotor network, cognitive control network and default pattern network [16], [17]. Structurally, IBD patients had reduced gray matter temporal gyrus, right anterior central gyrus, right auxiliary motor area, and right middle frontal gyrus compared with healthy controls. Compared with the healthy control group, a decrease was observed in the gray matter volume and cortical thickness of bilateral fusiform gyrus, infratemporal gyrus, right anterior central gyrus, right auxiliary motor area, and right middle frontal gyrus in IBD patients [18].

However, the above studies only focused on the regional changes of brain BOLD signal or gray matter structure, and few studies on IBD used multimodal data. Considering the complexity of the brain and the importance of functional and structural connections, we combined these two sources of information with tools of the emerging field of graph signal processing (GSP). The GSP was used to analyze the data in the irregular domain, which can be naturally represented as a graph. The data on the graph nodes are considered as a signal strongly dependent on the graph topology [19]. GSP has been applied to related research on medical images and has achieved good results [20].

In this paper, we proposed a new method for the analysis of IBD data using two different imaging modalities: fMRI and DKI. GSP technology was used to integrate the information of two kinds of data, and in comparison, with the traditional functional network analysis framework, the evaluation was carried out from different perspectives of the classification effect, which proved the superiority of multimodal data analysis. This method provides a new perspective for follow-up clinical auxiliary diagnosis.