Impaired interhemispheric synchrony in patients with iridocyclitis and classification using machine learning: an fMRI study

Abstract

Background: This study examined the interhemispheric integration function pattern in patients with iridocyclitis utilizing the voxel-mirrored homotopic connectivity (VMHC) technique. Additionally, we investigated the ability of VMHC results to distinguish patients with iridocyclitis from healthy controls (HCs), which may contribute to the development of objective biomarkers for early diagnosis and intervention in clinical set.

Methods: Twenty-six patients with iridocyclitis and twenty-six matched HCs, in terms of sex, age, and education level, underwent resting-state functional magnetic resonance imaging (fMRI) examinations. The study employed the voxel-mirrored homotopic connectivity (VMHC) technique to evaluate interhemispheric integration functional connectivity indices at a voxel-wise level. The diagnostic efficacy of VMHC was evaluated using a support vector machine (SVM) classifier, with classifier performance assessed through permutation test analysis. Furthermore, correlation analyses was conducted to investigate the associations between mean VMHC values in various brain regions and clinical features.

Results: Patients with iridocyclitis exhibited significantly reduced VMHC signal values in the bilateral inferior temporal gyrus, calcarine, middle temporal gyrus, and precuneus compared to HCs (voxel-level P < 0.01, Gaussian Random Field correction; cluster-level P < 0.05). Furthermore, the extracted resting-state zVMHC features effectively classified patients with iridocyclitis and HCs, achieving an area under the receiver operating characteristic curve (AUC) of 0.74 and an overall accuracy of 0.673 (P < 0.001, non-parametric permutation test).

Conclusion: Our findings reveal disrupted interhemispheric functional organization in patients with iridocyclitis, offering insight into the pathophysiological mechanisms associated with vision loss and cognitive dysfunction in this patient population. This study also highlights the potential of machine learning in ophthalmology and the importance of establishing objective biomarkers to address diagnostic heterogeneity.

Keywords: fMRI; interhemispheric integration; iridocyclitis; machine learning; support vector machine.

Figure 1

Typical ocular features in patients with iridocyclitis.

Figure 2

VMHC spatial maps at the group average level of HCs (A) and patients of iridocyclitis group (B) (voxel- level P<0.01, Gaussian random field correction, cluster-level P<0.05). The warm color tone suggests higher homotopic connectivity coefficients. The color bar represents the t-values. VMHC, voxel-mirrored homotopic connectivity; HC, healthy control; L, left; R, right.

Figure 3

(A) Altered VMHC values differences in the patient group compared with the control group. Blue-indigo color indicates reduced VMHC in iridocyclitis. Patients with iridocyclitis displayed significantly reduced VMHC signal values in the bilateral inferior temporal gyrus, calcarine, middle temporal gyrus, and precuneus compared with HCs. (voxel- level P<0.01, Gaussian random field correction, cluster-level P<0.05). VMHC, voxel-mirrored homotopic connectivity; HC, healthy control; L, left; R, right. (B) The mean values of changed VMHC between patients and healthy HCs.

Figure 4

Classification results using machine learning analysis based on zVMHC. (A) three-dimensional confusion matrices from SVM classifier; (B) function values of two groups (class 1: iridocyclitis; class 2: HC group); (C) the ROC curve of the SVM model, and the AUC was 0.74; (D) weight maps for the SVM classifier. VMHC, voxel-mirrored homotopic connectivity; SVM, support vector machine; ROC, receiver operating characteristic; AUC, area under the receiver operating characteristic curve; HC, healthy control.