Combining Anatomical and Functional MRI in Diagnosis of Drug-naive First Episode Schizophrenia

Wenting Ren,Peilin Lv,Wei Deng,Qizhu Wu,Xiaoqi Huang,Tao Li,Qiyong Gong,Su Lui

2011-01-01

Abstract:Introduction Although numerous morphometric and functional MRI studies have revealed anatomical and functional deficits mainly involving fronto-striato-parietal networks in patients with schizophrenia [1], it is still unclear whether the findings of MRI could be used for the diagnosis of this severe psychological disease, especially for drug-naive first episode schizophrenia (FES). Furthermore, increased evidences have been found to support the neural-network deficits in the core pathogenesis of schizophrenia [2]. Thus, a method taking account of multiple regional information and inter-regional correlations should be used for the distinction of FES patients from healthy controls. The present work aims to combine anatomic and functional MRI indiscriminate 100 drug-naïve FES from 100 healthy controls by utilizing a support vector machine (SVM) classification approach. Method The study was approved by the local ethical committee and written informed consent was obtained from all subjects. A total of 100 right-handed drug-naïve FES patients and 100 age, sex, height, weight, handedness and years of education matched normal controls were recruited in this study. The diagnoses of schizophrenia patients were determined using the structured clinical interview for DSM-IV patient edition. All patients were evaluated using the Global Assessment of Functioning Scale and clinical severity was evaluated using the PANSS which provides a total score, positive and negative symptom scores, and six indices of thought disturbance, activation, paranoia, depression, anergia and complementary. All subjects were scanned on a 3T Magnetic resonance scanner (EXCITE, General Electric, Milwaukee, USA) with a spoiled gradient recall sequence generating high-resolution T1-weighted volumetric 3D images and a gradient-echo echo-planar imaging sequence for resting-state fMRI. VBM8 in SPM8 was used to calculate the alterations of gray matter volume (GMV) in FES. The amplitude of low frequency (0.01–0.8 Hz) fluctuations (ALFF) of BOLD signals, which is thought to reflect spontaneous neural activity [3], were calculated by DPARSF software package(http://www.restfmri.net) to characterize regional cerebral function. SVM classification approach was used to calculate the sensitivity, specificity and accuracy for distinction of FES patients from healthy controls by fusing information from GMV and ALFF. This included three steps, i.e., firstly, a SVM based classifier ensemble is constructed using GMV; Secondly, ALFF were analyzed in SVM. Finally, the results obtained from the above two stages are combined into a single module using majority voting (Combined GMV-ALFF). The significant were set at P< 0.001 after permutation testing. Results The SVM classification between schizophrenia and controls using combined VBM and ALFF maps were show in the left of Figure 1. The classification yielded an accuracy of 83.5% (82% sensitivity, 85% specificity) with P< 0.001 after permutation testing. The most discriminating regions of GMV and ALFF were identified by setting the threshold to the top 30% of the weight vector scores (Figure 1) and these regions were listed as below (Table 1).