基于脑影像的精神分裂症定性诊断与定量分析的模式识别研究

Schizophrenia is a severe and disabling brain disorder which is diagnosed based on observed behavior and the person's self-reported experiences. Therefore, there is an urgent need for biomarkers that are both sensitive and specific for reliable and quantitative diagnosis. Multiparametric magnetic resonance images (MRIs), including structure MRI, functional MRI, and diffusion tensor MRI, have been being investigated as promising surrogate biomarkers. Although the potential of these biomarkers for quantitative diagnosis of schizophrenia has been demonstrated in many studies, most of the studies just focused on one or several different biomarkers and the effectiveness of these multiparametric measures has not been comprehensively evaluated. The evaluation of biomarkers in most studies has typically relied on independent comparison of individual measures derived from multiparametric MRI data and drawn conclusions at group level. However, no single measure carries information with sufficient sensitivity and specificity for quantitative diagnosis of schizophrenia, and the conclusions at group level cannot be easily extended for diagnosis of schizophrenia at individual level. These evaluation methods ignore the potential complementary information of multiple measures derived either from one individual biomarker or from multiple different biomarkers. Such complementary information of multiple measures has been demonstrated to be helpful to maximize the biomarkers' sensitivity and specificity for quantitative diagnosis of schizophrenia. This application aims to evaluate the effectiveness of imaging biomarkers for quantitative diagnosis of schizophrenia at individual level by using multivariate pattern recognition methods that maximize the sensitivity and specificity of the biomarkers. Specifically, we will develop a multivariate pattern classification system, including techniques for feature extraction from multiparametric MRI data, feature fusion, feature selection, and feature based classification techniques, for maximizing the MRI data's sensitivity and specificity for quantitative diagnosis of schizophrenia. These techniques will be integrated into a prototype computer aided diagnosis system of schizophrenia based on MRI data.

精神分裂症是一种严重的精神疾病，尚无可靠、客观的生物学诊断指标，其诊断主要依据患者的病史与临床症状表现。发展可靠、客观的精神分裂症早期诊断与预警的定量指标，对于早期识别与干预至关重要。大量研究表明，多参数磁共振脑影像有助于发现精神分裂症所导致的脑结构与脑功能病变。但是目前大部分脑影像学研究局限于组间差异统计分析，不能在个体水平上评测脑影像所反映的精神分裂症导致的脑异常。本项目突破传统的脑影像组间统计分析研究框架，拟在采集较大样本精神分裂症患者脑影像数据基础上，将模式识别方法与医学图像处理技术有机结合，发展准确、可靠的多参数磁共振脑影像特征提取、特征融合、以及特征选择新方法，研究高敏感度、高特异性的脑影像模式识别方法，建立基于脑影像的精神分裂症在个体水平上的辅助诊断原型系统，为疾病的准确诊断供客观依据。

精神分裂症是一种严重的精神疾病，尚无可靠、客观的生物学诊断指标，其诊断主要依据患者的病史与临床症状表现。发展可靠、客观的精神分裂症早期诊断与预警的定量指标，对于早期识别与干预至关重要。本项目将模式识别方法与医学图像处理技术有机结合，发展了准确、可靠的磁共振脑影像特征提取、特征融合、以及特征选择新方法，研究高敏感度、高特异性的脑影像模式识别方法。 在采集较大样本精神分裂症患者脑影像数据基础上，初步建立了基于脑影像的精神分裂症在个体水平上的辅助诊断算法，为精神分裂症的准确诊断提供了客观依据，并在此基础上进一步开展了精神分裂症患者家系研究及疗效评价研究。